responses.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "! pip install --quiet environs cyksuid toolz psycopg2-binary typing_json statsmodels\n",
    "! pip install --quiet backoff xxhash pyyaml geopandas dtreeviz facebook_business pystache\n",
    "! pip install --quiet bambi\n",
    "! pip install git+https://github.com/nandanrao/typedjson-python"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from environs import Env\n",
    "from adopt.malaria import get_db_conf\n",
    "\n",
    "env = Env()\n",
    "env.read_env('.env-bug', override=True)\n",
    "\n",
    "db_conf = get_db_conf(env)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import geopandas as gpd\n",
    "import pandas as pd\n",
    "from scipy.special import expit\n",
    "\n",
    "\n",
    "def get_dist_info(path):\n",
    "    geod = gpd.read_file(path)\n",
    "    district_info = geod.groupby('disthash').apply(lambda df: df.iloc[0][['disthash', 'tot_p']]).reset_index(drop=True)\n",
    "    return district_info\n",
    "\n",
    "dist_stats = get_dist_info('outs/cluster.shp')\n",
    "treatment_assignment = pd.read_csv('outs/ma-with-treatment.csv')[['disthash', 'treatment']]\n",
    "dist_info = pd.read_csv('outs/dist_info.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 672,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from mnm.prep import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 674,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "forms, form_keys = get_form_data(['malaria no more'], db_conf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "[f['shortcode'] for f in forms]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 852,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from mnm.prep import get_response_df_round_a\n",
    "\n",
    "dda = get_response_df_round_a(['malaria no more'], 'http://localhost:1323', db_conf)\n",
    "dda = dda \\\n",
    "    .merge(treatment_assignment, how='right', left_on='clusterid', right_on='disthash') \\\n",
    "    .merge(dist_info, on='clusterid')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "rbn.home.notna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 851,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# NOTE:\n",
    "# can't do this anymore, need raw df, same for costs.\n",
    "# dd[dd.wave == 0].to_csv('outs/individual-for-balance.csv', index=False)\n",
    "\n",
    "our_users = set(dda[dda.wave >= 1].userid.unique())\n",
    "dda = dda[dda.userid.isin(our_users)].reset_index(drop=True)\n",
    "dda['stratumid'] = dda.clusterid\n",
    "# dda.to_csv('outs/mnm-dda.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1368,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:root:Removing 6 possible test users\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:root:Removing 0 users for duplication\n"
     ]
    }
   ],
   "source": [
    "from mnm.prep import get_response_df\n",
    "\n",
    "ddb = get_response_df(['MNM', 'malaria no more round b 1-shot'], \n",
    "                      db_conf, \n",
    "                      ['stratumid'], \n",
    "                      'stratumid')\n",
    "\n",
    "# TODO: loses all the addition ind-effect folk, should include them!\n",
    "ddb = ddb \\\n",
    "    .merge(treatment_assignment, how='right', left_on='stratumid', right_on='disthash') \\\n",
    "    .merge(dist_info, left_on='stratumid', right_on='clusterid')\n",
    "\n",
    "# ddb = ddb[ddb.wave == '0']\n",
    "# ddb.to_csv('outs/mnm-ddb.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dda.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1379,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    9400\n",
       "1    5410\n",
       "2    2675\n",
       "3    2069\n",
       "9    1848\n",
       "4    1232\n",
       "5     844\n",
       "6     575\n",
       "7     204\n",
       "8      25\n",
       "Name: wave, dtype: int64"
      ]
     },
     "execution_count": 1379,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dda.wave.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1380,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.19659574468085106"
      ]
     },
     "execution_count": 1380,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "1848/9400"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1369,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ind_effect_users = ddb[ddb.shortcode.isin({'extraendhin'})].userid.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1370,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ied = ddb[ddb.userid.isin(ind_effect_users)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1371,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ups_users = ied[(ied.wave == '1') & (ied.week < 10)].userid.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1372,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ied = ied[~ied.userid.isin(ups_users)].reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1373,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ind_assignment = pd.read_csv('outs/ind-effect-with-assignment.csv')[['userid', 'ind_treatment']]\n",
    "ind_assignment['userid'] = ind_assignment.userid.astype(int).astype(str)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1374,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ied = ied.merge(ind_assignment)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1375,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "netusers = ied[(ied.hasmosquitonet == 'Yes')].userid.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1377,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>userid</th>\n",
       "      <th>stratumid</th>\n",
       "      <th>wave</th>\n",
       "      <th>language</th>\n",
       "      <th>week</th>\n",
       "      <th>month</th>\n",
       "      <th>survey_duration</th>\n",
       "      <th>survey_start_time</th>\n",
       "      <th>survey_end_time</th>\n",
       "      <th>shortcode</th>\n",
       "      <th>...</th>\n",
       "      <th>cost_per_message</th>\n",
       "      <th>CTR</th>\n",
       "      <th>CPM</th>\n",
       "      <th>high_CPM</th>\n",
       "      <th>low_CPM</th>\n",
       "      <th>avg_under_net</th>\n",
       "      <th>avg_long_sleeves</th>\n",
       "      <th>avg_malaria_now</th>\n",
       "      <th>tot_p</th>\n",
       "      <th>ind_treatment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3451582291634341</td>\n",
       "      <td>07853f76</td>\n",
       "      <td>0</td>\n",
       "      <td>hindi</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>95 days 04:20:48.152000</td>\n",
       "      <td>2021-01-07 05:19:59.041000+00:00</td>\n",
       "      <td>2021-04-12 09:40:47.193000+00:00</td>\n",
       "      <td>extrabasehin</td>\n",
       "      <td>...</td>\n",
       "      <td>11.111111</td>\n",
       "      <td>1.976214</td>\n",
       "      <td>59.168097</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>0.371429</td>\n",
       "      <td>0.685714</td>\n",
       "      <td>0.023891</td>\n",
       "      <td>3673889.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3451582291634341</td>\n",
       "      <td>07853f76</td>\n",
       "      <td>1</td>\n",
       "      <td>hindi</td>\n",
       "      <td>15</td>\n",
       "      <td>4</td>\n",
       "      <td>0 days 00:07:22.933000</td>\n",
       "      <td>2021-04-12 09:42:02.380000+00:00</td>\n",
       "      <td>2021-04-12 09:49:25.313000+00:00</td>\n",
       "      <td>extraendhin</td>\n",
       "      <td>...</td>\n",
       "      <td>11.111111</td>\n",
       "      <td>1.976214</td>\n",
       "      <td>59.168097</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>0.371429</td>\n",
       "      <td>0.685714</td>\n",
       "      <td>0.023891</td>\n",
       "      <td>3673889.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3544335325657230</td>\n",
       "      <td>07853f76</td>\n",
       "      <td>0</td>\n",
       "      <td>hindi</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>89 days 03:06:22.795000</td>\n",
       "      <td>2021-01-07 06:00:57.372000+00:00</td>\n",
       "      <td>2021-04-06 09:07:20.167000+00:00</td>\n",
       "      <td>extrabasehin</td>\n",
       "      <td>...</td>\n",
       "      <td>11.111111</td>\n",
       "      <td>1.976214</td>\n",
       "      <td>59.168097</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>0.371429</td>\n",
       "      <td>0.685714</td>\n",
       "      <td>0.023891</td>\n",
       "      <td>3673889.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3544335325657230</td>\n",
       "      <td>07853f76</td>\n",
       "      <td>1</td>\n",
       "      <td>hindi</td>\n",
       "      <td>14</td>\n",
       "      <td>4</td>\n",
       "      <td>5 days 13:45:33.378000</td>\n",
       "      <td>2021-04-06 14:58:51.514000+00:00</td>\n",
       "      <td>2021-04-12 04:44:24.892000+00:00</td>\n",
       "      <td>extraendhin</td>\n",
       "      <td>...</td>\n",
       "      <td>11.111111</td>\n",
       "      <td>1.976214</td>\n",
       "      <td>59.168097</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>0.371429</td>\n",
       "      <td>0.685714</td>\n",
       "      <td>0.023891</td>\n",
       "      <td>3673889.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3555534061199684</td>\n",
       "      <td>07853f76</td>\n",
       "      <td>0</td>\n",
       "      <td>hindi</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>94 days 00:33:40.070000</td>\n",
       "      <td>2021-01-08 08:49:48.167000+00:00</td>\n",
       "      <td>2021-04-12 09:23:28.237000+00:00</td>\n",
       "      <td>extrabasehin</td>\n",
       "      <td>...</td>\n",
       "      <td>11.111111</td>\n",
       "      <td>1.976214</td>\n",
       "      <td>59.168097</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>0.371429</td>\n",
       "      <td>0.685714</td>\n",
       "      <td>0.023891</td>\n",
       "      <td>3673889.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2797</th>\n",
       "      <td>4816349428437450</td>\n",
       "      <td>ff099f99</td>\n",
       "      <td>0</td>\n",
       "      <td>hindi</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>94 days 10:14:21.680000</td>\n",
       "      <td>2021-01-08 00:15:28.666000+00:00</td>\n",
       "      <td>2021-04-12 10:29:50.346000+00:00</td>\n",
       "      <td>extrabasehin</td>\n",
       "      <td>...</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>1.917183</td>\n",
       "      <td>71.270758</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>0.652174</td>\n",
       "      <td>0.586957</td>\n",
       "      <td>0.010791</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2798</th>\n",
       "      <td>4867033496701877</td>\n",
       "      <td>ff099f99</td>\n",
       "      <td>0</td>\n",
       "      <td>hindi</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>95 days 01:29:22.365000</td>\n",
       "      <td>2021-01-07 09:35:50.291000+00:00</td>\n",
       "      <td>2021-04-12 11:05:12.656000+00:00</td>\n",
       "      <td>extrabasehin</td>\n",
       "      <td>...</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>1.917183</td>\n",
       "      <td>71.270758</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>0.652174</td>\n",
       "      <td>0.586957</td>\n",
       "      <td>0.010791</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2799</th>\n",
       "      <td>4867033496701877</td>\n",
       "      <td>ff099f99</td>\n",
       "      <td>1</td>\n",
       "      <td>hindi</td>\n",
       "      <td>15</td>\n",
       "      <td>4</td>\n",
       "      <td>0 days 10:35:29.446000</td>\n",
       "      <td>2021-04-12 11:05:38.549000+00:00</td>\n",
       "      <td>2021-04-12 21:41:07.995000+00:00</td>\n",
       "      <td>extraendhin</td>\n",
       "      <td>...</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>1.917183</td>\n",
       "      <td>71.270758</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>0.652174</td>\n",
       "      <td>0.586957</td>\n",
       "      <td>0.010791</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2800</th>\n",
       "      <td>4950322035039336</td>\n",
       "      <td>ff099f99</td>\n",
       "      <td>0</td>\n",
       "      <td>hindi</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>93 days 19:19:59.555000</td>\n",
       "      <td>2021-01-08 14:44:58.591000+00:00</td>\n",
       "      <td>2021-04-12 10:04:58.146000+00:00</td>\n",
       "      <td>extrabasehin</td>\n",
       "      <td>...</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>1.917183</td>\n",
       "      <td>71.270758</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>0.652174</td>\n",
       "      <td>0.586957</td>\n",
       "      <td>0.010791</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2801</th>\n",
       "      <td>4950322035039336</td>\n",
       "      <td>ff099f99</td>\n",
       "      <td>1</td>\n",
       "      <td>hindi</td>\n",
       "      <td>15</td>\n",
       "      <td>4</td>\n",
       "      <td>0 days 00:03:52.479000</td>\n",
       "      <td>2021-04-12 10:05:48.460000+00:00</td>\n",
       "      <td>2021-04-12 10:09:40.939000+00:00</td>\n",
       "      <td>extraendhin</td>\n",
       "      <td>...</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>1.917183</td>\n",
       "      <td>71.270758</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>0.652174</td>\n",
       "      <td>0.586957</td>\n",
       "      <td>0.010791</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2802 rows × 78 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                userid stratumid wave language  week  month  \\\n",
       "0     3451582291634341  07853f76    0    hindi     1      1   \n",
       "1     3451582291634341  07853f76    1    hindi    15      4   \n",
       "2     3544335325657230  07853f76    0    hindi     1      1   \n",
       "3     3544335325657230  07853f76    1    hindi    14      4   \n",
       "4     3555534061199684  07853f76    0    hindi     1      1   \n",
       "...                ...       ...  ...      ...   ...    ...   \n",
       "2797  4816349428437450  ff099f99    0    hindi     1      1   \n",
       "2798  4867033496701877  ff099f99    0    hindi     1      1   \n",
       "2799  4867033496701877  ff099f99    1    hindi    15      4   \n",
       "2800  4950322035039336  ff099f99    0    hindi     1      1   \n",
       "2801  4950322035039336  ff099f99    1    hindi    15      4   \n",
       "\n",
       "             survey_duration                survey_start_time  \\\n",
       "0    95 days 04:20:48.152000 2021-01-07 05:19:59.041000+00:00   \n",
       "1     0 days 00:07:22.933000 2021-04-12 09:42:02.380000+00:00   \n",
       "2    89 days 03:06:22.795000 2021-01-07 06:00:57.372000+00:00   \n",
       "3     5 days 13:45:33.378000 2021-04-06 14:58:51.514000+00:00   \n",
       "4    94 days 00:33:40.070000 2021-01-08 08:49:48.167000+00:00   \n",
       "...                      ...                              ...   \n",
       "2797 94 days 10:14:21.680000 2021-01-08 00:15:28.666000+00:00   \n",
       "2798 95 days 01:29:22.365000 2021-01-07 09:35:50.291000+00:00   \n",
       "2799  0 days 10:35:29.446000 2021-04-12 11:05:38.549000+00:00   \n",
       "2800 93 days 19:19:59.555000 2021-01-08 14:44:58.591000+00:00   \n",
       "2801  0 days 00:03:52.479000 2021-04-12 10:05:48.460000+00:00   \n",
       "\n",
       "                      survey_end_time     shortcode  ... cost_per_message  \\\n",
       "0    2021-04-12 09:40:47.193000+00:00  extrabasehin  ...        11.111111   \n",
       "1    2021-04-12 09:49:25.313000+00:00   extraendhin  ...        11.111111   \n",
       "2    2021-04-06 09:07:20.167000+00:00  extrabasehin  ...        11.111111   \n",
       "3    2021-04-12 04:44:24.892000+00:00   extraendhin  ...        11.111111   \n",
       "4    2021-04-12 09:23:28.237000+00:00  extrabasehin  ...        11.111111   \n",
       "...                               ...           ...  ...              ...   \n",
       "2797 2021-04-12 10:29:50.346000+00:00  extrabasehin  ...        20.000000   \n",
       "2798 2021-04-12 11:05:12.656000+00:00  extrabasehin  ...        20.000000   \n",
       "2799 2021-04-12 21:41:07.995000+00:00   extraendhin  ...        20.000000   \n",
       "2800 2021-04-12 10:04:58.146000+00:00  extrabasehin  ...        20.000000   \n",
       "2801 2021-04-12 10:09:40.939000+00:00   extraendhin  ...        20.000000   \n",
       "\n",
       "           CTR        CPM high_CPM low_CPM avg_under_net avg_long_sleeves  \\\n",
       "0     1.976214  59.168097    False    True      0.371429         0.685714   \n",
       "1     1.976214  59.168097    False    True      0.371429         0.685714   \n",
       "2     1.976214  59.168097    False    True      0.371429         0.685714   \n",
       "3     1.976214  59.168097    False    True      0.371429         0.685714   \n",
       "4     1.976214  59.168097    False    True      0.371429         0.685714   \n",
       "...        ...        ...      ...     ...           ...              ...   \n",
       "2797  1.917183  71.270758     True   False      0.652174         0.586957   \n",
       "2798  1.917183  71.270758     True   False      0.652174         0.586957   \n",
       "2799  1.917183  71.270758     True   False      0.652174         0.586957   \n",
       "2800  1.917183  71.270758     True   False      0.652174         0.586957   \n",
       "2801  1.917183  71.270758     True   False      0.652174         0.586957   \n",
       "\n",
       "     avg_malaria_now      tot_p ind_treatment  \n",
       "0           0.023891  3673889.0             0  \n",
       "1           0.023891  3673889.0             0  \n",
       "2           0.023891  3673889.0             0  \n",
       "3           0.023891  3673889.0             0  \n",
       "4           0.023891  3673889.0             0  \n",
       "...              ...        ...           ...  \n",
       "2797        0.010791        NaN             0  \n",
       "2798        0.010791        NaN             1  \n",
       "2799        0.010791        NaN             1  \n",
       "2800        0.010791        NaN             0  \n",
       "2801        0.010791        NaN             0  \n",
       "\n",
       "[2802 rows x 78 columns]"
      ]
     },
     "execution_count": 1376,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "Male      1257\n",
       "Female     142\n",
       "Other        2\n",
       "Name: gender, dtype: int64"
      ]
     },
     "execution_count": 1377,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ied.gender.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1378,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "ename": "ConversionError",
     "evalue": "Failed to convert value(s) to axis units: 0               52\n1              NaN\n2       7920367458\n3              NaN\n4               22\n           ...    \n2797            25\n2798            27\n2799           NaN\n2800            24\n2801           NaN\nName: x, Length: 2802, dtype: object",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/axis.py\u001b[0m in \u001b[0;36mconvert_units\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m   1488\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1489\u001b[0;31m             \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconverter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvert\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munits\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1490\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/category.py\u001b[0m in \u001b[0;36mconvert\u001b[0;34m(value, unit, axis)\u001b[0m\n\u001b[1;32m     61\u001b[0m         \u001b[0;31m# force an update so it also does type checking\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 62\u001b[0;31m         \u001b[0munit\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     63\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvectorize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0munit\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_mapping\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__getitem__\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0motypes\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mfloat\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/category.py\u001b[0m in \u001b[0;36mupdate\u001b[0;34m(self, data)\u001b[0m\n\u001b[1;32m    219\u001b[0m             \u001b[0;31m# OrderedDict just iterates over unique values in data.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 220\u001b[0;31m             \u001b[0m_api\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheck_isinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mstr\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbytes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvalue\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mval\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    221\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mconvertible\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/_api/__init__.py\u001b[0m in \u001b[0;36mcheck_isinstance\u001b[0;34m(_types, **kwargs)\u001b[0m\n\u001b[1;32m     91\u001b[0m                 \u001b[0mnames\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"None\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 92\u001b[0;31m             raise TypeError(\n\u001b[0m\u001b[1;32m     93\u001b[0m                 \"{!r} must be an instance of {}, not a {}\".format(\n",
      "\u001b[0;31mTypeError\u001b[0m: 'value' must be an instance of str or bytes, not a float",
      "\nThe above exception was the direct cause of the following exception:\n",
      "\u001b[0;31mConversionError\u001b[0m                           Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1378-4a7d11a6f6d1>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0msns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdisplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mied\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/distributions.py\u001b[0m in \u001b[0;36mdisplot\u001b[0;34m(data, x, y, hue, row, col, weights, kind, rug, rug_kws, log_scale, legend, palette, hue_order, hue_norm, color, col_wrap, row_order, col_order, height, aspect, facet_kws, **kwargs)\u001b[0m\n\u001b[1;32m   2225\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2226\u001b[0m             \u001b[0m_assign_default_kwargs\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhist_kws\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_univariate_histogram\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhistplot\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2227\u001b[0;31m             \u001b[0mp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_univariate_histogram\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mhist_kws\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2228\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2229\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/distributions.py\u001b[0m in \u001b[0;36mplot_univariate_histogram\u001b[0;34m(self, multiple, element, fill, common_norm, common_bins, shrink, kde, kde_kws, color, legend, line_kws, estimate_kws, **plot_kws)\u001b[0m\n\u001b[1;32m    422\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    423\u001b[0m         \u001b[0;31m# First pass through the data to compute the histograms\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 424\u001b[0;31m         \u001b[0;32mfor\u001b[0m \u001b[0msub_vars\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msub_data\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0miter_data\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"hue\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfrom_comp_data\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    425\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    426\u001b[0m             \u001b[0;31m# Prepare the relevant data\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/_core.py\u001b[0m in \u001b[0;36miter_data\u001b[0;34m(self, grouping_vars, reverse, from_comp_data)\u001b[0m\n\u001b[1;32m    980\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    981\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mfrom_comp_data\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 982\u001b[0;31m             \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcomp_data\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    983\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    984\u001b[0m             \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_data\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/_core.py\u001b[0m in \u001b[0;36mcomp_data\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1049\u001b[0m                 \u001b[0maxis\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0max\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34mf\"{var}axis\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1050\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1051\u001b[0;31m                 \u001b[0mcomp_var\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvert_units\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_data\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mvar\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1052\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_scale\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"log\"\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1053\u001b[0m                     \u001b[0mcomp_var\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlog10\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcomp_var\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/axis.py\u001b[0m in \u001b[0;36mconvert_units\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m   1489\u001b[0m             \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconverter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvert\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munits\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1490\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1491\u001b[0;31m             raise munits.ConversionError('Failed to convert value(s) to axis '\n\u001b[0m\u001b[1;32m   1492\u001b[0m                                          f'units: {x!r}') from e\n\u001b[1;32m   1493\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mret\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mConversionError\u001b[0m: Failed to convert value(s) to axis units: 0               52\n1              NaN\n2       7920367458\n3              NaN\n4               22\n           ...    \n2797            25\n2798            27\n2799           NaN\n2800            24\n2801           NaN\nName: x, Length: 2802, dtype: object"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "ename": "ConversionError",
     "evalue": "Failed to convert value(s) to axis units: 0               52\n1              NaN\n2       7920367458\n3              NaN\n4               22\n           ...    \n2797            25\n2798            27\n2799           NaN\n2800            24\n2801           NaN\nName: x, Length: 2802, dtype: object",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/axis.py\u001b[0m in \u001b[0;36mconvert_units\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m   1488\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1489\u001b[0;31m             \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconverter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvert\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munits\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1490\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/category.py\u001b[0m in \u001b[0;36mconvert\u001b[0;34m(value, unit, axis)\u001b[0m\n\u001b[1;32m     61\u001b[0m         \u001b[0;31m# force an update so it also does type checking\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 62\u001b[0;31m         \u001b[0munit\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     63\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvectorize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0munit\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_mapping\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__getitem__\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0motypes\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mfloat\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/category.py\u001b[0m in \u001b[0;36mupdate\u001b[0;34m(self, data)\u001b[0m\n\u001b[1;32m    219\u001b[0m             \u001b[0;31m# OrderedDict just iterates over unique values in data.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 220\u001b[0;31m             \u001b[0m_api\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheck_isinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mstr\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbytes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvalue\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mval\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    221\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mconvertible\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/_api/__init__.py\u001b[0m in \u001b[0;36mcheck_isinstance\u001b[0;34m(_types, **kwargs)\u001b[0m\n\u001b[1;32m     91\u001b[0m                 \u001b[0mnames\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"None\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 92\u001b[0;31m             raise TypeError(\n\u001b[0m\u001b[1;32m     93\u001b[0m                 \"{!r} must be an instance of {}, not a {}\".format(\n",
      "\u001b[0;31mTypeError\u001b[0m: 'value' must be an instance of str or bytes, not a float",
      "\nThe above exception was the direct cause of the following exception:\n",
      "\u001b[0;31mConversionError\u001b[0m                           Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1378-4a7d11a6f6d1>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0msns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdisplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mied\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/distributions.py\u001b[0m in \u001b[0;36mdisplot\u001b[0;34m(data, x, y, hue, row, col, weights, kind, rug, rug_kws, log_scale, legend, palette, hue_order, hue_norm, color, col_wrap, row_order, col_order, height, aspect, facet_kws, **kwargs)\u001b[0m\n\u001b[1;32m   2225\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2226\u001b[0m             \u001b[0m_assign_default_kwargs\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhist_kws\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_univariate_histogram\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhistplot\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2227\u001b[0;31m             \u001b[0mp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_univariate_histogram\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mhist_kws\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2228\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2229\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/distributions.py\u001b[0m in \u001b[0;36mplot_univariate_histogram\u001b[0;34m(self, multiple, element, fill, common_norm, common_bins, shrink, kde, kde_kws, color, legend, line_kws, estimate_kws, **plot_kws)\u001b[0m\n\u001b[1;32m    422\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    423\u001b[0m         \u001b[0;31m# First pass through the data to compute the histograms\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 424\u001b[0;31m         \u001b[0;32mfor\u001b[0m \u001b[0msub_vars\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msub_data\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0miter_data\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"hue\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfrom_comp_data\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    425\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    426\u001b[0m             \u001b[0;31m# Prepare the relevant data\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/_core.py\u001b[0m in \u001b[0;36miter_data\u001b[0;34m(self, grouping_vars, reverse, from_comp_data)\u001b[0m\n\u001b[1;32m    980\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    981\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mfrom_comp_data\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 982\u001b[0;31m             \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcomp_data\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    983\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    984\u001b[0m             \u001b[0mdata\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_data\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/seaborn/_core.py\u001b[0m in \u001b[0;36mcomp_data\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1049\u001b[0m                 \u001b[0maxis\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0max\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34mf\"{var}axis\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1050\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1051\u001b[0;31m                 \u001b[0mcomp_var\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvert_units\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot_data\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mvar\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1052\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0maxis\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_scale\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"log\"\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1053\u001b[0m                     \u001b[0mcomp_var\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlog10\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcomp_var\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.9/site-packages/matplotlib/axis.py\u001b[0m in \u001b[0;36mconvert_units\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m   1489\u001b[0m             \u001b[0mret\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconverter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvert\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munits\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1490\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1491\u001b[0;31m             raise munits.ConversionError('Failed to convert value(s) to axis '\n\u001b[0m\u001b[1;32m   1492\u001b[0m                                          f'units: {x!r}') from e\n\u001b[1;32m   1493\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mret\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mConversionError\u001b[0m: Failed to convert value(s) to axis units: 0               52\n1              NaN\n2       7920367458\n3              NaN\n4               22\n           ...    \n2797            25\n2798            27\n2799           NaN\n2800            24\n2801           NaN\nName: x, Length: 2802, dtype: object"
     ]
    }
   ],
   "source": [
    "sns.displot(ied.age)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1014,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ind_treatment  treatment  sleepundernet\n",
       "0              0          1.0              0.673554\n",
       "                          0.0              0.326446\n",
       "               1          1.0              0.709251\n",
       "                          0.0              0.290749\n",
       "1              0          1.0              0.711027\n",
       "                          0.0              0.288973\n",
       "               1          1.0              0.736842\n",
       "                          0.0              0.263158\n",
       "Name: sleepundernet, dtype: float64"
      ]
     },
     "execution_count": 1014,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ied[(ied.wave == '1') & (ied.userid.isin(netusers))].groupby([\"ind_treatment\", \"treatment\"]).sleepundernet.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1053,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.535564\n",
      "         Iterations 5\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>       <td>admalaria</td>    <th>  No. Observations:  </th>  <td>   875</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>   871</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     3</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Tue, 13 Apr 2021</td> <th>  Pseudo R-squ.:     </th> <td>0.006247</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>16:30:06</td>     <th>  Log-Likelihood:    </th> <td> -468.62</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -471.56</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>    <th>  LLR p-value:       </th>  <td>0.1170</td> \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "             <td></td>                <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th>               <td>   -1.3808</td> <td>    0.165</td> <td>   -8.372</td> <td> 0.000</td> <td>   -1.704</td> <td>   -1.058</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>ind_treatment</th>           <td>    0.4617</td> <td>    0.218</td> <td>    2.119</td> <td> 0.034</td> <td>    0.035</td> <td>    0.889</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th>               <td>    0.0301</td> <td>    0.238</td> <td>    0.126</td> <td> 0.899</td> <td>   -0.436</td> <td>    0.496</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>ind_treatment:treatment</th> <td>   -0.3736</td> <td>    0.326</td> <td>   -1.144</td> <td> 0.253</td> <td>   -1.013</td> <td>    0.266</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:              admalaria   No. Observations:                  875\n",
       "Model:                          Logit   Df Residuals:                      871\n",
       "Method:                           MLE   Df Model:                            3\n",
       "Date:                Tue, 13 Apr 2021   Pseudo R-squ.:                0.006247\n",
       "Time:                        16:30:06   Log-Likelihood:                -468.62\n",
       "converged:                       True   LL-Null:                       -471.56\n",
       "Covariance Type:            nonrobust   LLR p-value:                    0.1170\n",
       "===========================================================================================\n",
       "                              coef    std err          z      P>|z|      [0.025      0.975]\n",
       "-------------------------------------------------------------------------------------------\n",
       "Intercept                  -1.3808      0.165     -8.372      0.000      -1.704      -1.058\n",
       "ind_treatment               0.4617      0.218      2.119      0.034       0.035       0.889\n",
       "treatment                   0.0301      0.238      0.126      0.899      -0.436       0.496\n",
       "ind_treatment:treatment    -0.3736      0.326     -1.144      0.253      -1.013       0.266\n",
       "===========================================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 1053,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smf.logit('admalaria ~ ind_treatment*treatment', ied[(ied.userid.isin(netusers)) & ied.sleepundernet.notna()]).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1365,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ied[(ied.userid.isin(netusers)) & ied.sleepundernet.notna()].to_csv('outs/mnm-ied.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1360,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\\begin{center}\n",
      "\\begin{tabular}{lclc}\n",
      "\\toprule\n",
      "\\textbf{Dep. Variable:}           &  sleepundernet   & \\textbf{  R-squared:         } &     0.002   \\\\\n",
      "\\textbf{Model:}                   &       OLS        & \\textbf{  Adj. R-squared:    } &    -0.001   \\\\\n",
      "\\textbf{Method:}                  &  Least Squares   & \\textbf{  F-statistic:       } &    0.7418   \\\\\n",
      "\\textbf{Date:}                    & Wed, 14 Apr 2021 & \\textbf{  Prob (F-statistic):} &    0.527    \\\\\n",
      "\\textbf{Time:}                    &     20:08:14     & \\textbf{  Log-Likelihood:    } &   -593.68   \\\\\n",
      "\\textbf{No. Observations:}        &         941      & \\textbf{  AIC:               } &     1195.   \\\\\n",
      "\\textbf{Df Residuals:}            &         937      & \\textbf{  BIC:               } &     1215.   \\\\\n",
      "\\textbf{Df Model:}                &           3      & \\textbf{                     } &             \\\\\n",
      "\\bottomrule\n",
      "\\end{tabular}\n",
      "\\begin{tabular}{lcccccc}\n",
      "                                  & \\textbf{coef} & \\textbf{std err} & \\textbf{t} & \\textbf{P$> |$t$|$} & \\textbf{[0.025} & \\textbf{0.975]}  \\\\\n",
      "\\midrule\n",
      "\\textbf{Intercept}                &       0.6736  &        0.029     &    22.993  &         0.000        &        0.616    &        0.731     \\\\\n",
      "\\textbf{ind\\_treatment}           &       0.0375  &        0.041     &     0.923  &         0.356        &       -0.042    &        0.117     \\\\\n",
      "\\textbf{treatment}                &       0.0357  &        0.042     &     0.848  &         0.397        &       -0.047    &        0.118     \\\\\n",
      "\\textbf{ind\\_treatment:treatment} &      -0.0099  &        0.060     &    -0.166  &         0.868        &       -0.127    &        0.107     \\\\\n",
      "\\bottomrule\n",
      "\\end{tabular}\n",
      "\\begin{tabular}{lclc}\n",
      "\\textbf{Omnibus:}       & 643.256 & \\textbf{  Durbin-Watson:     } &    1.944  \\\\\n",
      "\\textbf{Prob(Omnibus):} &   0.000 & \\textbf{  Jarque-Bera (JB):  } &  182.055  \\\\\n",
      "\\textbf{Skew:}          &  -0.905 & \\textbf{  Prob(JB):          } & 2.93e-40  \\\\\n",
      "\\textbf{Kurtosis:}      &   1.830 & \\textbf{  Cond. No.          } &     6.66  \\\\\n",
      "\\bottomrule\n",
      "\\end{tabular}\n",
      "%\\caption{OLS Regression Results}\n",
      "\\end{center}\n",
      "\n",
      "Notes: \\newline\n",
      " [1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n"
     ]
    }
   ],
   "source": [
    "print(smf.ols('sleepundernet ~ ind_treatment*treatment', ied[(ied.userid.isin(netusers)) & ied.sleepundernet.notna()]).fit().summary().as_latex())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1359,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.6735532026572099,\n",
       " 0.7110316958114002,\n",
       " 0.7092408682678242,\n",
       " 0.7368383401414632)"
      ]
     },
     "execution_count": 1359,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.7243), expit(0.7243 + 0.1761), expit(0.7243 + 0.1674), expit(0.7243 + 0.1761 + 0.1674 - 0.0382)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1006,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ind_treatment  treatment  admalaria\n",
       "0              0          0.0          0.799127\n",
       "                          1.0          0.200873\n",
       "               1          0.0          0.794258\n",
       "                          1.0          0.205742\n",
       "1              0          0.0          0.714876\n",
       "                          1.0          0.285124\n",
       "               1          0.0          0.779487\n",
       "                          1.0          0.220513\n",
       "Name: admalaria, dtype: float64"
      ]
     },
     "execution_count": 1006,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ied[(ied.wave == '1') & (ied.userid.isin(netusers))].groupby([\"ind_treatment\", \"treatment\"]).admalaria.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1028,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = Model(ied[(ied.userid.isin(netusers)) & ied.sleepundernet.notna()].reset_index(drop=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1031,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:bambi:Modeling the probability that sleepundernet==1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:Auto-assigning NUTS sampler...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:Initializing NUTS using jitter+adapt_diag...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:Multiprocess sampling (4 chains in 4 jobs)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:NUTS: [treatment:ind_treatment, ind_treatment, treatment, Intercept]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='0' class='' max='24000' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      \n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:Sampling 4 chains for 4_000 tune and 2_000 draw iterations (16_000 + 8_000 draws total) took 26 seconds.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:pymc3:The acceptance probability does not match the target. It is 0.8838984138847404, but should be close to 0.8. Try to increase the number of tuning steps.\n"
     ]
    }
   ],
   "source": [
    "results = model.fit('sleepundernet ~ treatment*ind_treatment', family='bernoulli', draws=2_000, tune=4_000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1034,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([<AxesSubplot:title={'center':'treatment'}>,\n",
       "       <AxesSubplot:title={'center':'ind_treatment'}>,\n",
       "       <AxesSubplot:title={'center':'treatment:ind_treatment'}>],\n",
       "      dtype=object)"
      ]
     },
     "execution_count": 1034,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1490.4x331.2 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "az.plot_posterior(results, var_names=['treatment','ind_treatment', 'treatment:ind_treatment'], ref_val=0, rope=[-1, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1050,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0795"
      ]
     },
     "execution_count": 1050,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "((results.posterior.treatment.values < 0) & (results.posterior.ind_treatment.values < 0)).mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1051,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.69375"
      ]
     },
     "execution_count": 1051,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "((results.posterior.treatment.values > 0) & (results.posterior.ind_treatment.values > 0)).mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 517,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def remove_spammers(df):\n",
    "    df = df[(df.answer_time_median.dt.total_seconds()/df.answer_time_min.dt.total_seconds()) >= 1.5]\n",
    "    return df.reset_index(drop=True)\n",
    "\n",
    "ddb = remove_spammers(ddb[ddb.dwelling.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 856,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dda = remove_spammers(dda)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "def as_type(series, type_):\n",
    "    return series.astype(type_)\n",
    "\n",
    "def binarize(series, pos):\n",
    "    s = as_type((series == pos), int)\n",
    "    s[series.isna()] = np.nan\n",
    "    return s\n",
    "\n",
    "def binarize_cols(df, cols):\n",
    "    df = df.copy()\n",
    "    for col, target in cols:\n",
    "        df[col] = binarize(df[col], target)\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 857,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "cols = [\n",
    "    ('malaria2weeks', 'Yes'), \n",
    "    ('fever2weeks', 'Yes'), \n",
    "    ('seekhelpfever', 'Yes'),\n",
    "    ('longsleeves', 'Yes'),\n",
    "    ('sleepundernet', 'Yes'),\n",
    "    ('timeseekhelpfever', 'Less than 24 hours')\n",
    "]\n",
    "\n",
    "dda = binarize_cols(dda, cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 342,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dda['stratumid'] = dda.clusterid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 489,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Yes           1134\n",
       "No              78\n",
       "Don’t know       1\n",
       "Name: seekhelpfever, dtype: int64"
      ]
     },
     "execution_count": 489,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[ddb.dwelling.notna() & ddb.seekhelpfever.notna()].seekhelpfever.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 488,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "No            156\n",
       "Yes            69\n",
       "Don’t know     16\n",
       "Name: seekhelpfever, dtype: int64"
      ]
     },
     "execution_count": 488,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[ddb.dwelling.isna() & ddb.seekhelpfever.notna()].seekhelpfever.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1407,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "cols = [\n",
    "    ('malaria4months', 'Yes'), \n",
    "    ('fever4months', 'Yes'), \n",
    "    ('seekhelpfever', 'Yes'),\n",
    "    ('longsleeves', 'Yes'),\n",
    "    ('sleepundernet', 'Yes'),\n",
    "    ('admnm', 'Yes'),\n",
    "    ('admalaria', 'Yes'),\n",
    "    ('knowmnm', 'Yes'),\n",
    "    ('testmalaria', 'Yes'),\n",
    "]\n",
    "\n",
    "ddb = binarize_cols(ddb, cols)\n",
    "ied = binarize_cols(ied, cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 620,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0    0.639175\n",
       "1.0    0.360825\n",
       "Name: fever4months, dtype: float64"
      ]
     },
     "execution_count": 620,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[(ddb.malaria4months == 1)].fever4months.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 621,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0    0.510264\n",
       "1.0    0.489736\n",
       "Name: fever2weeks, dtype: float64"
      ]
     },
     "execution_count": 621,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dda[dda.malaria2weeks == 1].fever2weeks.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 629,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Yes           206\n",
       "No            114\n",
       "Don’t know     21\n",
       "Name: testmalaria, dtype: int64"
      ]
     },
     "execution_count": 629,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dda[(dda.malaria2weeks == 1) & (dda.fever2weeks == 0)].testmalaria.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# TODO: build model for malaria incidence that considers\n",
    "# the selection bias - the fact that only some people \n",
    "# get tested, so there is a probability of getting tested,\n",
    "# and then try to infer effects on malaria incidence."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "vpre = get_pair_delta(dda, 'seekhelpfever', ['stratumid', 'treatment'], 'pair_id', pair_lookup, dif('seekhelpfever'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import statsmodels.formula.api as smf\n",
    "\n",
    "pp = vpre[['delta', 'pair_id']].rename(columns={'delta': 'pre_delta'}).merge(v).groupby('pair_id').head(1).reset_index(drop=True)\n",
    "\n",
    "smf.ols('delta ~ 1 + pre_delta', pp).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "pp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "from toolz import curry\n",
    "from scipy.stats import rankdata\n",
    "\n",
    "@curry\n",
    "def dif(var, df):\n",
    "    d = df[df.treatment == 1][var].iloc[0] - df[df.treatment == 0][var].iloc[0]\n",
    "    rd = df[df.treatment == 1]['rank'].iloc[0] - df[df.treatment == 0]['rank'].iloc[0]\n",
    "    return df.assign(delta = d, count = df['count'].sum(), rank_delta = rd)\n",
    "\n",
    "def center(a):\n",
    "    return a - np.mean(a)\n",
    "\n",
    "def get_pair_delta(df, var, cluster_groups, pair_id, pair_lookup, fn):\n",
    "    v = df[df[var].notnull()] \\\n",
    "        .groupby(cluster_groups) \\\n",
    "        .agg({var: 'mean', 'userid': 'count'}) \\\n",
    "        .rename(columns={'userid': 'count'}) \\\n",
    "        .pipe(lambda df: df.assign(rank = center(rankdata(df[var])))) \\\n",
    "        .reset_index() \\\n",
    "        .merge(pair_lookup) \\\n",
    "        .groupby(pair_id) \\\n",
    "        .apply(fn)[['delta', 'count', 'rank_delta', pair_id]] \\\n",
    "        .reset_index(drop=True) \\\n",
    "        .sort_values('count')\n",
    "\n",
    "    return v\n",
    "\n",
    "\n",
    "\n",
    "# sns.distplot(v['delta'], bins=20)\n",
    "\n",
    "# np.mean(v.delta)\n",
    "\n",
    "pair_lookup = pd.read_csv('outs/ma-with-pair-id.csv')[['pair_id', 'disthash']].rename(columns={'disthash': 'stratumid'})\n",
    "v = get_pair_delta(ddb, 'seekhelpfever', ['stratumid', 'treatment'], 'pair_id', pair_lookup, dif('seekhelpfever'))\n",
    "# v\n",
    "\n",
    "sns.distplot(v['delta'], bins=20)\n",
    "# np.mean(v.delta)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 522,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dd = da.rename(columns={'seekhelpfever': 'pre_seekhelpfever'}).merge(d[['seekhelpfever', 'stratumid']], on='stratumid')\n",
    "dd['delta'] = dd.seekhelpfever - dd.pre_seekhelpfever"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "smf.ols('delta ~ treatment', dd.merge(dist_info, left_on='stratumid', right_on='clusterid')).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "d = ddb[ddb.seekhelpfever.notna() & ddb.dwelling.notna()].groupby(['stratumid', 'treatment']).agg({'seekhelpfever': 'mean', 'userid': 'count'}).reset_index()\n",
    "\n",
    "smf.ols('seekhelpfever ~ treatment', d).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sns.histplot(dd, x='delta', hue=\"treatment\", bins=30, kde=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 892,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:xlabel='seekhelpfever', ylabel='Count'>"
      ]
     },
     "execution_count": 892,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAEGCAYAAABiq/5QAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Z1A+gAAAACXBIWXMAAAsTAAALEwEAmpwYAAArj0lEQVR4nO3deZwU9Z3/8denr+m5hzk4nGEYQEURvAKJSqKJRvQnWd1o4rHGaNQ1Zk2ykRxryGaNiUncNboaY5IlJjGX5jCaO954Bg9QFAUUEWaY4RoG5r76+P7+qAYGHOgGprtnpt/Px6MfVV1dU99P18B7qr9d9S1zziEiIrnDl+0CREQksxT8IiI5RsEvIpJjFPwiIjlGwS8ikmMC2S4gFZWVla6uri7bZYiIjChLly7d6pyr2nP5iAj+uro6lixZku0yRERGFDOrH2y5unpERHKMgl9EJMco+EVEcsyI6OMfTCQSobGxkd7e3myXkjbhcJiamhqCwWC2SxGRUWTEBn9jYyPFxcXU1dVhZtkuZ8g552hpaaGxsZHJkydnuxwRGUVGbFdPb28vFRUVozL0AcyMioqKUf2JRkSyY8QGPzBqQ3+H0f7+RCQ7RnTwi4jI/hs1wd/a2sr3v//9Ybet4dCOiMhAoz74o9HokG1rqCn4RXLTxNpJmFlKj4m1k4a8/RF7Vs+errvuOtasWcOxxx5LMBgkHA4zZswYVq1axcqVK7nuuut44okn6Ovr45prruGTn/wknZ2dnHPOOWzfvp1IJMKNN97IOeecs9u2Tj/9dObNm8f1119PWVkZy5cv5/zzz2fmzJncfvvt9PT08Ic//IGpU6fS3NzM1VdfTUNDAwC33XYbc+bM4Wtf+xoNDQ28/fbbNDQ08LnPfY7Pfvaz72jn5ptvzvJeFJFMaFzfwK0Pv5HSuvPnThvy9kdN8N9000289tprLFu2jCeeeIJ58+bx2muvMXnyZBYuXEhpaSkvvvgifX19zJkzh7lz5zJx4kQeeOABSkpK2Lp1KyeccAJnn332btsCeOKJJ3jllVdYuXIl5eXlTJkyhSuvvJIXXniB22+/nTvuuIPbbruNf//3f+faa6/lve99Lw0NDZxxxhmsXLkSgFWrVrFo0SI6OjqYNm0an/rUp97RjohIJoya4N/Tu9/97p3nvz/88MO8+uqr3HfffQC0tbWxevVqampqWLBgAU899RQ+n4+mpiY2b9486PZmz57NhAkTAJg6dSpz584FYObMmSxatAiARx99lBUrVuz8mfb2djo7OwGYN28eeXl55OXlMXbs2L22IyKSbqM2+AsLC3fOO+e44447OOOMM3Zb5+6776a5uZmlS5cSDAapq6vb63nzeXl5O+d9Pt/O5z6fb+f3CPF4nOeee45wOLzPn/f7/Qf03YOIyFAYNV/uFhcX09HRMehrZ5xxBj/4wQ+IRCIAvPnmm3R1ddHW1sbYsWMJBoMsWrSI+vr6pNval7lz53LHHXfsfJ6sC+dA2xERORijJvgrKiqYM2cOM2bM4Itf/OJur1155ZVMnz6d448/nhkzZvDJT36SaDTKxRdfzJIlS5g5cyY///nPOeKII5Jua1+++93vsmTJEo4++mimT5/OD3/4wwOuWUQkXcw5l+0akpo1a5bb80YsK1eu5Mgjj8xSRZmTK+9TJJeY2X6d1XOgOW1mS51zs/ZcPmqO+EVEJDUKfhGRHKPgFxHJMQp+EZEck7bgN7OfmNkWM3ttwLKbzWyVmb1qZg+YWVm62hcRkcGl84j/buDMPZY9Asxwzh0NvAl8OY3ti4jIINIW/M65p4Bteyx72Dm345LV54CadLW/p/0ZDW+oRsx78MEHmTZtGoceeig33XRTBt6liEhy2Ryy4XLgN5lqbH9Gw0tFshHzYrEY11xzDY888gg1NTXMnj2bs88+m+nTpw9ZDSIiByIrX+6a2VeAKPCrfaxzlZktMbMlzc3NmStuiLzwwgsceuihTJkyhVAoxIUXXsgf//jHbJclIpL54Dezy4APARe7fVyO5pxb6Jyb5ZybVVVVlbH6hkpTUxMTJ07c+bympoampqYsViQi4sloV4+ZnQl8CTjFOdedybZFRMSTztM57wUWA9PMrNHMrgC+BxQDj5jZMjPb9yhmI1h1dTXr16/f+byxsZHq6uosViQi4knbEb9z7qJBFv84Xe0NN7Nnz2b16tWsXbuW6upqfv3rX3PPPfdkuywRkdF7I5Y91UysHdJ7V9ZMrN3n64FAgO9973ucccYZxGIxLr/8co466qgha19E5EDlTPCvb6jPeJtnnXUWZ511VsbbFRHZF43VIyKSYxT8IiI5RsEvIpJjFPwiIjlGwS8ikmMU/CIiOSZngr+utmZIh2Wuq00+ovTll1/O2LFjmTFjRgbeoYhIanLmPP769U24x781ZNuzUxckXeeyyy7j05/+NB//+MeHrF0RkYOVM0f82XDyySdTXl6e7TJERHaj4BcRyTEKfhGRHKPgFxHJMQp+EZEckzNn9UyaWJ3SmTj7s71kLrroIp544gm2bt1KTU0NN9xwA1dcccWQ1SAiciByJvjXNTRmvM177703422KiCSjrh4RkRyj4BcRyTEjOvidc9kuIa1G+/sTkewYscEfDodpaWkZteHonKOlpYVwOJztUkRklBmxX+7W1NTQ2NhIc3NztktJm3A4TE1N8sHgRET2x4gN/mAwyOTJk7NdhojIiJO2rh4z+4mZbTGz1wYsKzezR8xsdWI6Jl3ti4jI4NLZx383cOYey64DHnPOHQY8lnguIiIZlLbgd849BWzbY/E5wM8S8z8D/jld7YuIyOAyfVbPOOfcxsT8JmBchtsXEcl5Wfty1znnzGyv52Ka2VXAVQC1tbUZq0tEJBNuuOGGrLWd6eDfbGYTnHMbzWwCsGVvKzrnFgILAWbNmjU6T9YXkZx1/aWnpbTe/GfvGfK2M93V8yfg0sT8pcAfM9y+iEjOS+fpnPcCi4FpZtZoZlcANwGnm9lq4IOJ5yIikkFp6+pxzl20l5dS+3wjIiJpMWLH6hERkQOj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEck5XgN7Nrzex1M3vNzO41s3A26hARyUUZD34zqwY+C8xyzs0A/MCFma5DRCRXZaurJwDkm1kAKAA2ZKkOEZGck/Hgd841Ad8BGoCNQJtz7uE91zOzq8xsiZktaW5uznSZInKQJtZOwsxSekysnZTtcnNKINMNmtkY4BxgMtAK/M7MPuac++XA9ZxzC4GFALNmzXKZrlNEDk7j+gZuffiNlNadP3damquRgbLR1fNBYK1zrtk5FwHuB07KQh0iIjkpG8HfAJxgZgVmZsBpwMos1CEikpOy0cf/PHAf8BKwPFHDwkzXISKSqzLexw/gnLseuD4bbYuI5DpduSsikmNSCn4zm5PKMhERGf5SPeK/I8VlIiIyzO2zj9/MTsQ71bLKzOYPeKkEb6gFEREZYZJ9uRsCihLrFQ9Y3g58JF1FiYhI+uwz+J1zTwJPmtndzrn6DNUkIiJplOrpnHlmthCoG/gzzrlT01GUiIikT6rB/zvgh8BdQCx95YiISLqlGvxR59wP0lqJiIhkRKqnc/7ZzP7NzCaYWfmOR1orExGRtEj1iP/SxPSLA5Y5YMrQliMiIumWUvA75yanuxAREcmMlILfzD4+2HLn3M+HthwREUm3VLt6Zg+YD+ONof8SoOAXERlhUu3q+czA52ZWBvw6HQWJiEh6HeiwzF1498wVEZERJtU+/j/jncUD3uBsRwK/TVdRIiKSPqn28X9nwHwUqHfONaahHhERSbOUunoSg7WtwhuhcwzQn86iREQkfVK9A9f5wAvAR4HzgefNTMMyi4iMQKl29XwFmO2c2wJgZlXAo8B96SpMRETSI9Wzenw7Qj+hZT9+VkREhpFUj/gfNLOHgHsTzy8A/paekkREJJ2S3XP3UGCcc+6LZnYu8N7ES4uBXx1oo4kLwO4CZuCdJnq5c27xgW5PRERSl+yI/zbgywDOufuB+wHMbGbitX86wHZvBx50zn3EzEJAwQFuR0RE9lOy4B/nnFu+50Ln3HIzqzuQBs2sFDgZuCyxrX50eqiISMYk+4K2bB+v5R9gm5OBZuCnZvaymd1lZoV7rmRmV5nZEjNb0tzcfIBNyXBVV1uDmSV91NXWZLtUkVEn2RH/EjP7V+fcjwYuNLMrgaUH0ebxwGecc8+b2e3AdcBXB67knFsILASYNWuWe8dWZESrX9+Ee/xbSdezUxdkoBqR3JIs+D8HPGBmF7Mr6GcBIeDDB9hmI9DonHs+8fw+vOAXEZEM2GfwO+c2AyeZ2QfwzsAB+Ktz7vEDbdA5t8nM1pvZNOfcG3hj+6840O2JiMj+SXU8/kXAoiFs9zPArxJn9LwNfGIIty0iIvuQ6gVcQ8o5twyvy0hERDJMwy6IiOQYBb+ISI5R8IuI5BgFv4hIjlHwi4jkGAW/iEiOUfCLiOQYBb+ISI5R8IuI5BgFv4hIjsnKkA0ikhtuuOGGbJcgg1Dwi0jaXH/paSmtN//Ze9JciQykrh4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEck7XgNzO/mb1sZn/JVg0iIrkom0f8/w6szGL7IiI5KSvBb2Y1wDzgrmy0LyKSy7J1xH8b8CUgvrcVzOwqM1tiZkuam5szVpiIZIH5MLOkj4m1k7Jd6aiQ8fH4zexDwBbn3FIze//e1nPOLQQWAsyaNctlpjoRyQoX59aH30i62vy50zJQzOiXjSP+OcDZZrYO+DVwqpn9Mgt1iIjkpIwHv3Puy865GudcHXAh8Lhz7mOZrkNEJFfpPH4RkRyT1XvuOueeAJ7IZg0iIrlGR/wiIjlGwS8ikmMU/CIiOUbBLyKSYxT8IiI5RsEvIpJjFPwiIjkmq+fxi0ju6Yz62NgbZHskQHfUR3fMR9V5/8Xfl28kGPCRH/RTWZRHZVGIMQUhfD7LdsmjjoJfRNIq7qCxJ8SbnWHqe0J0Rv07X8vzxSnwx/EXlbOls49INE5PJEY8MSxj0G9MqihkamUhdZWFWXoHo4+CX0TSwl9UwTMtRazsCNMd8xOyOJMK+plQ2sWEcITKUJRAorN5/o2f40uJ0Tljcce2rn62dvaxobWHt7d28daWTnwGlR/6Akvrt3F87RjM9EngQCn4RQSAibWTaFzfkHS9mom1rG+o3+vra5o7+b8n11B99V281BpgckEfRxR3MLmgb2fQ74vfZ1QV51FVnMeRE0o41Tk2t/fxxqYOXuqdzXk/WEzfprdoX/xbut9cDAwyarvPD/FY8rYCQaKR/uRFjTIKfhEBoHF9w0GNib+1s4/bHn2Te19YT9BvdCx7iM9++ARKg8kDeF/MjPGlYcaXhvnT/A9y2f89xUsFQfLGL6CyKMR7Jlcwtapwt08A8+dO49aFP07+Xq664qBqG6kU/CJyUPqjcX78zFruXPQWPZEYH3tPLZ857TCqis+i9PzZQ9qWi/Qys6aUo6pLeHNTB8+v3cZfl29kQmmYkw+vYnxJeEjbG60U/CJywJbWb2fB/ct5Y3MHHzxyLF8+60imVhWlvV2fGUdMKOHwccWs2NjOP9a08JsX13Pk+GJOmlqZ9vZHOgW/iOy3nv4Y3/77Sn7xXD3jS8L86OOzOH36uIzX4fMZM6pLOWxcEUvWbeflhlZWb+mk9KQLicZJ6TuFXKTgF5H9Ehw7mQ/d8TRrmru47KQ6vnDGNIryshsleQE/cw6tZEZ1Kc+8tZXo+z7Gz9fHeF9FB4cW9qETgHanv4cikhLnHC81bGfCJbfS0Rvll1e8h6+dfVTWQ3+g0vwg82ZOYNM915Hni/O3zWX8fsMYmvuGT43DgfaGiCTV1RflkRWbqd/WTc/apTx4w1cpLwxlu6y96lv/GhfVbOP19nz+sa2IexvLOaqkhxPLOynwD3L6Z47REb+I7NParV386vkGGlt7+MC0Kprvv3FYh/4OPoOZpT1cWruVY0q7WdGez8/qK1m8rZC+2PDo+3EOemNGS7+frX0BmvsCtPT7016fjvhFZFDRWJxn32phWWMrlUUhzjuqmoqivGyXtd/CfscplZ3MLOlh8bYiXthexCttBRxX2o0vvyRjdcTjjjXNnby4bjsV8+bzi4YKOqI+Im7w4++wL05JMEZe9fQhr0XBLyLv0NLZx99f30RLZz/HTixjztQKAv6R3UFQHooxb3wbW/q6eG5bEc9tL6L6Uz/lP+57lUtPqmP6IUP7R6AvGuO1pjZeXLedJeu2saR+O63dEQDy646jLBhlUkGM4kCcwkAMAwyIOeiI+mmP+mmL+IlHeoa0LlDwi8gAzjmWN7Xx1OqthPw+zjnmkFE3ONrYvChnT2ilpd/PD/78In8syOc3S9Zz6NgizpoxnrlHjefICSX493NU0C0dvSxvbGNp/XaWrNvOssZW+qNxAKZUFXLG9PHMqhvD7LpyJlcVMT+FK4sBnt6ydr/fYzIKfhEBwJdfwl9e3cjbW7uYVFHA6UeOo3AYnbEz1CpCMbY9dCdr7r+VP7+ygb8t38T3Fr3Fdx9/i8KQn5k1pcysLmVCaT5VxXmUF4aIO0ckFqc3EmdDaw+N23to2NbN6xva2NzeB0AgcW3BpSdOYlZdObMmjRl2XWQZ/62a2UTg58A4vNGVFjrnbs90HSKyyzOrtzLhE3dQ39LNyYdVcuzEspwZ/bKsIMQlJ9ZxyYl1NHf08fTqZpatb+WV9a387B/19Mfie/3ZwpCfmjEFnDilghnV3h+Ko2vKyA/59/ozw0E2/pxHgc87514ys2JgqZk94pxbkYVaRHJaXzTGLQ+/ycKn3ibe18UFs4+jqnh4HZ1mUlVxHuceX8O5x9cAXtdXW0+E5o4+tnX14/cZQb+PUMDHhNIwpfnBEfkHMuPB75zbCGxMzHeY2UqgGlDwi2TQig3tzP/tMlZt6uBjJ9TyrY+eR9UFr2a7rH3yG+T3byMcbScY7yEY6yWQmAbjvQTivZiLcc3sIMd2PoURx5zDcMTNTww/cQsQS8z/8xEBeOsxCJdCXgnkFUO4BIIFYIaZUVYQoqxg+J++uj+y2oFnZnXAccDzg7x2FXAVQG1tbWYLExliqY51D8nHuz9Ysbhj4VNvc+sjb1BWEOKnn5jNB6aN5ZvRPm644YaUtjHUR7kleVDRvYaivi0U9W+huG8LRf2bKezfSn6kjfxoK+FIG9f+Vwm8eEbS7Z12Vj60PZB0vXkXFMAvz33H8mjcsbXbsbnTsanT0UmY8z7+b1A0DkomQFkdlNVCYSUjcTyIrAW/mRUBvwc+55xr3/N159xCYCHArFmzdKmdjGipjnUPex/vfig0tHTz+d8t48V12/l/M8bzzQ/P3O1irOsvPS15fc/eg3v8W0nXs1MX7PY8FO+hPLqFsmgzY6LNlEa3Mia6lbLoVq69rgRevnDnug6jO1hOZ6iKnmAZbeFqeoKl/OHen3PKOf9Cn6+AfgsRtRARyyPiCxJJPI/j4z/nX8vX//d2wEc8caKkjzg+F8VPDL+L4ncx/vfr/8myH34Son0Q6/em0T4CsT7GR7oZ39/FMf1drFvfBM//H8T6dn+TwULvD0BZLZRPgappMPZIqDoC8suS7qNsyUrwm1kQL/R/5Zy7Pxs1iOQS5xy/fnE9N/5lBT4zbj3/GD58XHV6+qfjMehu4V9mBnhv25+piGykMrqRkljrrnow2v1jaA1U8kbBcfzigcf4wKdvoTNvHB2hsXSFKon7gu/Y9NefvItbL35f0hJaehx9vuSnob6yOQ6lE5OuN/kLC3Dxduhrh7YmaK2H7fXQ2rBrft3TEOne9UNF42HsETD2KDjkWDjkeO+Pgy/710Nk46weA34MrHTO3Zrp9kVyzdqtXXz5/ld57u1tnDilgu+cfwzVZflDs3EXh+4W6NgIHZu8aecWcDF+dW4Bsc4n2RYYR1NoKq8EJ7AtMI7tgSraAxXEbFf83LL471TfMHdoakoXM++7gHApjBvkatp4HNrWQ/Mq77ElMV3yE4gmLsLKK4VDjuHbp+VxaM8rNIWm0OMvzuz7IDtH/HOAS4DlZrYssWyBc+5vWahFZNSKxOIsfOptbn9sNXkBH98+dyYXzJqIbz8vTBooHOvinGkBWPN4IuQ3e10kAP6Qd5RbPQuKx3HUZ+7hqht/SNyG96mNQ8bngzGTvMfhA76HiEVh6xvQ9BJseBk2vMT8E0OEtt0NQEtgLE2hqTTmTaUpbyqd/rK0l5qNs3qewbsyWUTSZGn9dr7ywHJWberg/80Yzw1nH8XYA7gtYX6sg5r+NVT3raGmbw1V0Y186sICaFoKRWNh3AwonuA9Cip2+6JzRXM8d0J/X/wBGHeU9zj+EgCKA8bdt3+Vmr41VPe9zbSelzm6ezEArf4KGvIOpyF8OA15h6elpNF7WZ5IDtrS3stNf1/F/S83Mb4kzMJL3sXco8an/PMFsXZq+tZQ07+Gmr63qIhuBiBiITaE6ng2/1iu++59PP2T68Gn+DhQ/THYFKpjU6iOJcWnYS5OZWTDzv0+reclju5eTBzjwcOHfj/rNycyCvRH4/z02bV897HVRGKOaz4wlX97/6HJh1xoa4L6Z2HdM6y6ppBpm673tmd5NIUms7JgFo2hQ9kcqiGe6JN/puE3Cv0h5sxHc6iG5lANLxedgs/FGN9fz6S+N3hxw++HvD399kRGMOccD762if956A3Wbu3ig0eO5asfms6kir2c0dLaAOuehfpnvOn2xABgeaW82RJn85QP05g3lS3Bapy6abImbn425E1hQ94UNnXeN+TbV/CLjFDPvd3Ct/++ilfWt3LY2KKdF2Lt5JwX7OueTRzVPwttiYvIwmUwaQ68+1+96fiZnO0PcOupH8jKe5HMUvCLjDArNrTznYff4PFVW5hQGuZ/PnI05x1fgx8HW1Z6IV+/GBoWQ3uT90MFFV7An/Rpbzp2+rA4n1yyQ8EvMkK81tTG7Y+t5pEVmykJB7jujMO4bHIb4aYH4DeJoO/Z7q1cNB4mneiFfN37vCtKR+DQApIeCn6RYW55Yxu3P/Ymj67cQnEQPjdlA58IPEjpP/4BTyauFC2fCkfMg9qTvMAfM1lBL3ul4BcZjnpaeXnZEr63uIXHthRRYt1cG/grl/keonRjn3fl6HGXeCFfexIUj8t2xTKCKPhFssgf72dMzzoqu9ZQ2f0WY7rWMO7aUzj/63fxgjuSUuDzxY9y6aE9lNS+G6qvhvFHQ6gg26XLCKbgF0m3/i6OqPQxedvTlPWup6yn0Zv2NlLSuxEfMXpdkN/HT2Zh7HzqQ1VUhyN89ZggF7z/NIrKLsj2O5BRZtQHf11tDfXrm5KuN2liNesaGkdN28Oe+d4xbO/e1ktFxva1cxCLQF8H9Gzj3DNOpq9jK+X5RkW+UVVoVBcb1cU+qkuMQ4p9lIWNldcUwcr5ALRFA7zdXcjS7kJW9E7n8eD7qC85nniwgGjnNrY/fjP1q57mHy7Olfuqxef3RsJMwh8IEo30H/h7zpBU7wUwpPbj32EqI5mOlP/Loz7469c3HdDY4SO97WEtHqc4GOeW228hFO8j5HoJxXsJuT5C8V78RPG7KD4X46EHfguLvu0NBBbr90I31g8u5oUw3q0a/uu4Fi5fMCfxdODtG9xuk3seWwa/u8wbVTKe2IaLJebjie0mXot0Q6QnMe3dNe92he39/wQwoNvFfBAq8h55Rd4dnULFXHzL3/jQJ79Mq7+SXl8hzf1BXm3PZ1VHPlFn1BX0cXzZNmrCET5/55Mpjd0/f+40bl344+TrXXVF0nWGg1TvBTCkXDzlfTia/i+P+uCXDOnv9kZrbG+C9o3Q1QzdW6Frqzdsb3dLYn4r9LbR/uUS2JT8CO+U08Lw5E3eEAH+UOIRHDBkgHcUdvqUgDcm+p52O0ozZh/ig02veQHt84P5vXV2zg9YXlAJwXzvNny7TfO9QM8v58xzL+LBWz7lLQvkQyBv0LNp7ln+J44L1rGmK49X2grY0BvCb45pRb0cX9ZFRSj5kbvIUFHwS3LxuBfq29d6l/y3NSUCfkPi0bjr/PGBfAEvPAsrvQuIJhzjzeeP4fNfuYEzz/84/ZZHvy9MxPLoS0yjFiSGd1/Uz19zDX39saQXG9Wa4R5PfrR1+KkLcG7Jge6Jd3hozflQcsg+19nYG6R0zr/wk/pKumN+SgNR3lfRwfTiHsJ+3VxOMk/BL55Ib+JOQutg21ov5HdMt9e/85ZzBRVQUg2lNTDx3V74ldZ40+JDvIAPl+71XPJbF3+Fmk+cmLSs/hgj8grTvrjx6JYSfts0hqe2FlM6ZxpVeRGOKWmnrqBfp9hLVin4c0n3tt0Dfds6L+i3r/WO3Af2jQcLoXwyVB7u3VRizGTvedkkL/CD+z+2ey5Y0R7mt03l/GFjGa2RABPC/Xxmyhb+40sL+NzNyfuIRTJBwT+axOPQseGdR+w7pr1tu69fNM4L9Mknw5i6XeE+ZrJ3xK7D0pTUd4f4y6ZS/ryxjFWd+YQsztxx7ZxfvY05FZ34DT7ftjnbZYrspOAfSZzz+tLbGr1Ha8O+u2R8Ae9G0uWTofpdu0K9fLIX9KHkN6OWwa3d2sUjKzYx/uO3csrT3l2S3lXWxQ1HNHHOhFbK9GWtDGMK/uGkvxs6N3lfnu4I97b1A+YbIdK1+8/srUtmzGQv9P36FQ+FSCzOi+u28fjKLTy+agtvb931e1hw+AbmjW+jOj+SxQpFUqdUSLdYBKK90N+VeHR608iu5298uhC+VQP9He/8+cIq70vTqsPh0NO8+Z2Pid7r6pIZcn3RGK82tvH82y08v3YbS+u3090fI+T3ccLUCi49qY5TjxhLbcWHuOoy9d3LyKLgT8Y5iPZ5V2r2d3jT3nbobfW+LO3ZnnjsmN+1vGtBMTxzy+Db9ed5XS2hQl7aGOfwsz7m3by6aByUVnuhXnKId364pFVvJMYbmzp4fUM7r29o4/UN7azc2E5fNA7AtHHFfORdNZw0tZL3HVaZ/HaGIsPc6P4X3N9FdbF5Fw7F+iDa701j/V6YD5jec24+/Oqj0NfphXtfe2LaAfEkH+F9QSgoh/wxkF/u9Z8fchx3PvRjvnjxXAiEvUG1gkWQV+h1z/iDO3/8ovkLuPC+m9K7L3JUX9zY1h+gpT9Ac1+Awhmncueit6hv6WJdSzcNLd1sau/duX5xOMBRh5RwyQmTmD25nHfXlTOmMJTFdyAy9EZ38D/0FRrnF8OSu/a+ji8A/jzedYgPOrd4V2SWTfSmOx6hIsgrGbCsaFfI54/xjtwH6W750oe/zxe/nPxc9ZEkEovT3hOhLfFo743umu+J0N0fpTcSpzcS86bRGH2RGH3ROH2RODHniDvH+Etu4d7GcnDeSaR7u4xpwifu4Mzbnkpa14RP3MGZz44n4ozumI+uqI/umJ+o2/33UjlvPjc/9AZVxXlMKi/gpEMrmFReyOHjijjqkFImluenNCaLyEiWleA3szOB2wE/cJdzLj2HuzM/ypXXf5+7/uNC71L/QN6AaR4EQjsHAZt26gKcezItZQwn8bijsz9Ke0+E9p4o7b1eYLf3RncGenvvriBv7xkQ7L0Ruvv3fbaK32eEAz7CQT/hoJ+8gI+8oJ9w0EdewEfI58cM4r0dFPi9rhQDDDfoVxUNrZuoLT8h6ft6qXUTtQXlBM1REIhT6I+T7/emY0JRKhOPE664ibbN6ykIje5jHpF9yfi/fjPzA3cCpwONwItm9ifn3Iohb6xuDj9+OcJdY48cks055xLTXUeozrkB87DjmXOAP0hvzPZ43Uu3gUe5FipgW1c/0VicSNx505gjGo8TjTkisTjReGKaWN4bidPdH6OnP0pPJObNR2L09HuP7sR8d390t4Dv6IvikowSUJwXoCQ/SEl+kNL8AJMqCijND1K6c1mQkvzArmXhXa+Fg/6U9qVdeQLXnZ58cKzHH/gmC++/Mel6P7p0Ngs/k/xL1mjrJoW+5Lxs/A94N/CWc+5tADP7NXAOMOTB/42/rGDitfdx5KOhnWG3W7fCgACv/cIfmLrgb97iPcL8QE36wgMc8Wjy9Wqv/S3Hf+ORA28oIRTwURDykx/0k5+YFoT8TCgNM218cSKkE6Ee9sLbm3rPi8MBisMBAv6RN0SCiKTO3MEk24E0aPYR4Ezn3JWJ55cA73HOfXqP9a4Crko8nQYkH6c2MyqBrdkuIgnVePCGe30w/Gsc7vXB6K9xknOuas+Fw/Yzr3NuIbAw23XsycyWOOdmZbuOfVGNB2+41wfDv8bhXh/kbo3Z+EzfBEwc8LwmsUxERDIgG8H/InCYmU02sxBwIfCnLNQhIpKTMt7V45yLmtmngYfwTuf8iXPu9UzXcRCGXffTIFTjwRvu9cHwr3G41wc5WmPGv9wVEZHs0nl7IiI5RsEvIpJjFPx7YWZnmtkbZvaWmV03yOtXm9lyM1tmZs+Y2fThVuOA9c4zM2dmGT1tLYV9eJmZNSf24TIzuzKT9aVSY2Kd881shZm9bmb3DKf6zOx/B+y/N82sNZP1pVhjrZktMrOXzexVMztrGNY4ycweS9T3hJnVZLi+n5jZFjN7bS+vm5l9N1H/q2Z2/EE16JzTY48H3pfOa4ApQAh4BZi+xzolA+bPBh4cbjUm1isGngKeA2YNp/qAy4DvDfPf82HAy8CYxPOxw6m+Pdb/DN7JEsNtHy4EPpWYnw6sG4Y1/g64NDF/KvCLDNd4MnA88NpeXj8L+Dve0FYnAM8fTHs64h/czmElnHP9wI5hJXZyzrUPeFrI3geYTJekNSZ8A/hvoHeQ19Ip1fqyKZUa/xW40zm3HcA5t2WY1TfQRcC9Galsl1RqdEBJYr4U2JDB+iC1GqcDjyfmFw3yelo5554Ctu1jlXOAnzvPc0CZmU040PYU/IOrBtYPeN6YWLYbM7vGzNYA/wN8NkO17ZC0xsTHwYnOub9msrCElPYhcF7io+t9ZjZxkNfTKZUaDwcON7Nnzey5xMiymZLqPsTMJgGT2RVemZJKjV8DPmZmjcDf8D6ZZFIqNb4CnJuY/zBQbGYVGagtVSn/W0iFgv8gOOfudM5NBf4D+M9s1zOQmfmAW4HPZ7uWffgzUOecOxp4BPhZlusZTACvu+f9eEfUPzKzsmwWtBcXAvc554bjXd4vAu52ztXgdVn8IvHvczj5AnCKmb0MnII3msBw3JdDYrjt/OFif4eV+DXwz+ksaBDJaiwGZgBPmNk6vH7BP2XwC96k+9A51+Kc60s8vQt4V4Zq2yGV33Mj8CfnXMQ5txZ4E+8PwXCpb4cLyXw3D6RW4xXAbwGcc4uBMN7AY5mSyr/FDc65c51zxwFfSSxrzViFyQ3tUDeZ/AJjpDzwjvLexvvovOPLoKP2WOewAfP/BCwZbjXusf4TZPbL3VT24YQB8x8Gnhtu+xA4E/hZYr4S7+N2xXCpL7HeEcA6EhdkDsN9+HfgssT8kXh9/BmrNcUaKwFfYv6bwNezsC/r2PuXu/PY/cvdFw6qrUy/uZHywPtI+ibe2QBfSSz7OnB2Yv524HVgGd6XQXsN3WzVuMe6GQ3+FPfhtxP78JXEPjxiuO3DxH+0W/HuF7EcuHA41Zd4/jXgpkzvu/3Yh9OBZxO/52XA3GFY40eA1Yl17gLyMlzfvcBGIIL3KfMK4Grg6gH/Du9M1L/8YP8va8gGEZEcoz5+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgl5xlZneb2UcGWb7OzFK+wMjM3m9mf0lhvZsTI3zevL+1igyljN96USSHXQWUuzQNq2BmAedcNB3bltFFR/wyIplZoZn91cxeMbPXzOwCM3uXmT1pZkvN7KEdoxea2VQzezCx/GkzO2KQ7X0j8QnAn1j0GTN7KXHPhSMGtPkTM3shMbb8O0ZwNLOvmdkvzGyxma02s39NLP8TUAQsTdRaZWa/N7MXE485ZuZLfNooG7C91WY2brD192jvWeAXQ7ybZZTSEb+MVGcCG5xz8wDMrBTvkvZznHPNZnYB3qX3l+ONB3+1c261mb0H+D7emOskfvZmvLGNPuGcc2YGsNU5d7yZ/RveAF5X4o3h8rhz7vJEOL9gZo8OUtvReJfVFwIvm9lfnXNnm1mnc+7YRJv3AP/rnHvGzGqBh5xzR5rZH/GGr/hpotZ659zmwdbHG/4AvCtj3+uc6zn43Sq5QMEvI9Vy4BYz+2/gL8B2vEHpHkkEtx/YaGZFwEnA7xLLAfIGbOereDe1uGqP7d+fmC5l13C9c4GzzewLiedhoHaQ2v6YCOEeM1uENx78H/ZY54PA9AE1lSRq/Q3wX8BP8QZe+02S9cEbRE6hLylT8MuI5Jx7M3G/gbOAG/HGoX/dOXfiwPXMrARo3XGkPYgXgXeZWblzbuCNMHaMGhpj1/8TA85zzr2xRxvj9iwvyXPwullPcM7tdoMcM1sMHGpmVXgjvt6YZH2ArsHfmsjg1McvI5KZHQJ0O+d+CdwMvAeoMrMTE68Hzewo590pba2ZfTSx3MzsmAGbehC4CfirmRUnafYhvL5/S2zruL2sd46ZhRM38ng/3h+XPT3MgBuSmNmxAM4bPOsBvIHhVjrnWva1vsiBUPDLSDUTr499GXA9XvfIR4D/NrMdo0CelFj3YuCKxPLXeedtNH8H/AjvfgX5+2jzG0AQeNXMXk88H8yreKONPgd8wzk32K0GPwvMMu/uYyvwRmLc4TfAx9jVzZNsfZH9otE5RYaQmX0N6HTOfSfbtYjsjY74RURyjI74RURyjI74RURyjIJfRCTHKPhFRHKMgl9EJMco+EVEcsz/Bx1qnaBRGV0EAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "da = dda[(dda.week <= 34) & dda.seekhelpfever.notna()].groupby(['stratumid', 'treatment']).agg({'seekhelpfever': 'mean', 'userid': 'count'}).reset_index()\n",
    "sns.histplot(da, x='seekhelpfever', hue=\"treatment\", bins=30, kde=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 870,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>OLS Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>      <td>seekhelpfever</td>  <th>  R-squared:         </th> <td>   0.055</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                   <td>OLS</td>       <th>  Adj. R-squared:    </th> <td>   0.043</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>             <td>Least Squares</td>  <th>  F-statistic:       </th> <td>   4.573</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>             <td>Tue, 13 Apr 2021</td> <th>  Prob (F-statistic):</th>  <td>0.0356</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                 <td>13:57:15</td>     <th>  Log-Likelihood:    </th> <td>  30.077</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Observations:</th>      <td>    80</td>      <th>  AIC:               </th> <td>  -56.15</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Residuals:</th>          <td>    78</td>      <th>  BIC:               </th> <td>  -51.39</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Model:</th>              <td>     1</td>      <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>nonrobust</td>    <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>t</th>      <th>P>|t|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    0.8599</td> <td>    0.027</td> <td>   32.322</td> <td> 0.000</td> <td>    0.807</td> <td>    0.913</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>   -0.0805</td> <td>    0.038</td> <td>   -2.138</td> <td> 0.036</td> <td>   -0.155</td> <td>   -0.006</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "  <th>Omnibus:</th>       <td> 9.802</td> <th>  Durbin-Watson:     </th> <td>   2.355</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Prob(Omnibus):</th> <td> 0.007</td> <th>  Jarque-Bera (JB):  </th> <td>   9.752</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Skew:</th>          <td>-0.827</td> <th>  Prob(JB):          </th> <td> 0.00763</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Kurtosis:</th>      <td> 3.439</td> <th>  Cond. No.          </th> <td>    2.62</td>\n",
       "</tr>\n",
       "</table><br/><br/>Notes:<br/>[1] Standard Errors assume that the covariance matrix of the errors is correctly specified."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                            OLS Regression Results                            \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   R-squared:                       0.055\n",
       "Model:                            OLS   Adj. R-squared:                  0.043\n",
       "Method:                 Least Squares   F-statistic:                     4.573\n",
       "Date:                Tue, 13 Apr 2021   Prob (F-statistic):             0.0356\n",
       "Time:                        13:57:15   Log-Likelihood:                 30.077\n",
       "No. Observations:                  80   AIC:                            -56.15\n",
       "Df Residuals:                      78   BIC:                            -51.39\n",
       "Df Model:                           1                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      0.8599      0.027     32.322      0.000       0.807       0.913\n",
       "treatment     -0.0805      0.038     -2.138      0.036      -0.155      -0.006\n",
       "==============================================================================\n",
       "Omnibus:                        9.802   Durbin-Watson:                   2.355\n",
       "Prob(Omnibus):                  0.007   Jarque-Bera (JB):                9.752\n",
       "Skew:                          -0.827   Prob(JB):                      0.00763\n",
       "Kurtosis:                       3.439   Cond. No.                         2.62\n",
       "==============================================================================\n",
       "\n",
       "Notes:\n",
       "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n",
       "\"\"\""
      ]
     },
     "execution_count": 870,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smf.ols('seekhelpfever ~ treatment', da).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 896,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:xlabel='seekhelpfever', ylabel='Count'>"
      ]
     },
     "execution_count": 896,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAEGCAYAAABiq/5QAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Z1A+gAAAACXBIWXMAAAsTAAALEwEAmpwYAAArj0lEQVR4nO3deZwU9Z3/8denr+m5hzk4nGEYQEURvAKJSqKJRvQnWd1o4rHGaNQ1Zk2ykRxryGaNiUncNboaY5IlJjGX5jCaO954Bg9QFAUUEWaY4RoG5r76+P7+qAYGHOgGprtnpt/Px6MfVV1dU99P18B7qr9d9S1zziEiIrnDl+0CREQksxT8IiI5RsEvIpJjFPwiIjlGwS8ikmMC2S4gFZWVla6uri7bZYiIjChLly7d6pyr2nP5iAj+uro6lixZku0yRERGFDOrH2y5unpERHKMgl9EJMco+EVEcsyI6OMfTCQSobGxkd7e3myXkjbhcJiamhqCwWC2SxGRUWTEBn9jYyPFxcXU1dVhZtkuZ8g552hpaaGxsZHJkydnuxwRGUVGbFdPb28vFRUVozL0AcyMioqKUf2JRkSyY8QGPzBqQ3+H0f7+RCQ7RnTwi4jI/hs1wd/a2sr3v//9Ybet4dCOiMhAoz74o9HokG1rqCn4RXLTxNpJmFlKj4m1k4a8/RF7Vs+errvuOtasWcOxxx5LMBgkHA4zZswYVq1axcqVK7nuuut44okn6Ovr45prruGTn/wknZ2dnHPOOWzfvp1IJMKNN97IOeecs9u2Tj/9dObNm8f1119PWVkZy5cv5/zzz2fmzJncfvvt9PT08Ic//IGpU6fS3NzM1VdfTUNDAwC33XYbc+bM4Wtf+xoNDQ28/fbbNDQ08LnPfY7Pfvaz72jn5ptvzvJeFJFMaFzfwK0Pv5HSuvPnThvy9kdN8N9000289tprLFu2jCeeeIJ58+bx2muvMXnyZBYuXEhpaSkvvvgifX19zJkzh7lz5zJx4kQeeOABSkpK2Lp1KyeccAJnn332btsCeOKJJ3jllVdYuXIl5eXlTJkyhSuvvJIXXniB22+/nTvuuIPbbruNf//3f+faa6/lve99Lw0NDZxxxhmsXLkSgFWrVrFo0SI6OjqYNm0an/rUp97RjohIJoya4N/Tu9/97p3nvz/88MO8+uqr3HfffQC0tbWxevVqampqWLBgAU899RQ+n4+mpiY2b9486PZmz57NhAkTAJg6dSpz584FYObMmSxatAiARx99lBUrVuz8mfb2djo7OwGYN28eeXl55OXlMXbs2L22IyKSbqM2+AsLC3fOO+e44447OOOMM3Zb5+6776a5uZmlS5cSDAapq6vb63nzeXl5O+d9Pt/O5z6fb+f3CPF4nOeee45wOLzPn/f7/Qf03YOIyFAYNV/uFhcX09HRMehrZ5xxBj/4wQ+IRCIAvPnmm3R1ddHW1sbYsWMJBoMsWrSI+vr6pNval7lz53LHHXfsfJ6sC+dA2xERORijJvgrKiqYM2cOM2bM4Itf/OJur1155ZVMnz6d448/nhkzZvDJT36SaDTKxRdfzJIlS5g5cyY///nPOeKII5Jua1+++93vsmTJEo4++mimT5/OD3/4wwOuWUQkXcw5l+0akpo1a5bb80YsK1eu5Mgjj8xSRZmTK+9TJJeY2X6d1XOgOW1mS51zs/ZcPmqO+EVEJDUKfhGRHKPgFxHJMQp+EZEck7bgN7OfmNkWM3ttwLKbzWyVmb1qZg+YWVm62hcRkcGl84j/buDMPZY9Asxwzh0NvAl8OY3ti4jIINIW/M65p4Bteyx72Dm345LV54CadLW/p/0ZDW+oRsx78MEHmTZtGoceeig33XRTBt6liEhy2Ryy4XLgN5lqbH9Gw0tFshHzYrEY11xzDY888gg1NTXMnj2bs88+m+nTpw9ZDSIiByIrX+6a2VeAKPCrfaxzlZktMbMlzc3NmStuiLzwwgsceuihTJkyhVAoxIUXXsgf//jHbJclIpL54Dezy4APARe7fVyO5pxb6Jyb5ZybVVVVlbH6hkpTUxMTJ07c+bympoampqYsViQi4sloV4+ZnQl8CTjFOdedybZFRMSTztM57wUWA9PMrNHMrgC+BxQDj5jZMjPb9yhmI1h1dTXr16/f+byxsZHq6uosViQi4knbEb9z7qJBFv84Xe0NN7Nnz2b16tWsXbuW6upqfv3rX3PPPfdkuywRkdF7I5Y91UysHdJ7V9ZMrN3n64FAgO9973ucccYZxGIxLr/8co466qgha19E5EDlTPCvb6jPeJtnnXUWZ511VsbbFRHZF43VIyKSYxT8IiI5RsEvIpJjFPwiIjlGwS8ikmMU/CIiOSZngr+utmZIh2Wuq00+ovTll1/O2LFjmTFjRgbeoYhIanLmPP769U24x781ZNuzUxckXeeyyy7j05/+NB//+MeHrF0RkYOVM0f82XDyySdTXl6e7TJERHaj4BcRyTEKfhGRHKPgFxHJMQp+EZEckzNn9UyaWJ3SmTj7s71kLrroIp544gm2bt1KTU0NN9xwA1dcccWQ1SAiciByJvjXNTRmvM177703422KiCSjrh4RkRyj4BcRyTEjOvidc9kuIa1G+/sTkewYscEfDodpaWkZteHonKOlpYVwOJztUkRklBmxX+7W1NTQ2NhIc3NztktJm3A4TE1N8sHgRET2x4gN/mAwyOTJk7NdhojIiJO2rh4z+4mZbTGz1wYsKzezR8xsdWI6Jl3ti4jI4NLZx383cOYey64DHnPOHQY8lnguIiIZlLbgd849BWzbY/E5wM8S8z8D/jld7YuIyOAyfVbPOOfcxsT8JmBchtsXEcl5Wfty1znnzGyv52Ka2VXAVQC1tbUZq0tEJBNuuOGGrLWd6eDfbGYTnHMbzWwCsGVvKzrnFgILAWbNmjU6T9YXkZx1/aWnpbTe/GfvGfK2M93V8yfg0sT8pcAfM9y+iEjOS+fpnPcCi4FpZtZoZlcANwGnm9lq4IOJ5yIikkFp6+pxzl20l5dS+3wjIiJpMWLH6hERkQOj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEck5XgN7Nrzex1M3vNzO41s3A26hARyUUZD34zqwY+C8xyzs0A/MCFma5DRCRXZaurJwDkm1kAKAA2ZKkOEZGck/Hgd841Ad8BGoCNQJtz7uE91zOzq8xsiZktaW5uznSZInKQJtZOwsxSekysnZTtcnNKINMNmtkY4BxgMtAK/M7MPuac++XA9ZxzC4GFALNmzXKZrlNEDk7j+gZuffiNlNadP3damquRgbLR1fNBYK1zrtk5FwHuB07KQh0iIjkpG8HfAJxgZgVmZsBpwMos1CEikpOy0cf/PHAf8BKwPFHDwkzXISKSqzLexw/gnLseuD4bbYuI5DpduSsikmNSCn4zm5PKMhERGf5SPeK/I8VlIiIyzO2zj9/MTsQ71bLKzOYPeKkEb6gFEREZYZJ9uRsCihLrFQ9Y3g58JF1FiYhI+uwz+J1zTwJPmtndzrn6DNUkIiJplOrpnHlmthCoG/gzzrlT01GUiIikT6rB/zvgh8BdQCx95YiISLqlGvxR59wP0lqJiIhkRKqnc/7ZzP7NzCaYWfmOR1orExGRtEj1iP/SxPSLA5Y5YMrQliMiIumWUvA75yanuxAREcmMlILfzD4+2HLn3M+HthwREUm3VLt6Zg+YD+ONof8SoOAXERlhUu3q+czA52ZWBvw6HQWJiEh6HeiwzF1498wVEZERJtU+/j/jncUD3uBsRwK/TVdRIiKSPqn28X9nwHwUqHfONaahHhERSbOUunoSg7WtwhuhcwzQn86iREQkfVK9A9f5wAvAR4HzgefNTMMyi4iMQKl29XwFmO2c2wJgZlXAo8B96SpMRETSI9Wzenw7Qj+hZT9+VkREhpFUj/gfNLOHgHsTzy8A/paekkREJJ2S3XP3UGCcc+6LZnYu8N7ES4uBXx1oo4kLwO4CZuCdJnq5c27xgW5PRERSl+yI/zbgywDOufuB+wHMbGbitX86wHZvBx50zn3EzEJAwQFuR0RE9lOy4B/nnFu+50Ln3HIzqzuQBs2sFDgZuCyxrX50eqiISMYk+4K2bB+v5R9gm5OBZuCnZvaymd1lZoV7rmRmV5nZEjNb0tzcfIBNyXBVV1uDmSV91NXWZLtUkVEn2RH/EjP7V+fcjwYuNLMrgaUH0ebxwGecc8+b2e3AdcBXB67knFsILASYNWuWe8dWZESrX9+Ee/xbSdezUxdkoBqR3JIs+D8HPGBmF7Mr6GcBIeDDB9hmI9DonHs+8fw+vOAXEZEM2GfwO+c2AyeZ2QfwzsAB+Ktz7vEDbdA5t8nM1pvZNOfcG3hj+6840O2JiMj+SXU8/kXAoiFs9zPArxJn9LwNfGIIty0iIvuQ6gVcQ8o5twyvy0hERDJMwy6IiOQYBb+ISI5R8IuI5BgFv4hIjlHwi4jkGAW/iEiOUfCLiOQYBb+ISI5R8IuI5BgFv4hIjsnKkA0ikhtuuOGGbJcgg1Dwi0jaXH/paSmtN//Ze9JciQykrh4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgFxHJMQp+EZEck7XgNzO/mb1sZn/JVg0iIrkom0f8/w6szGL7IiI5KSvBb2Y1wDzgrmy0LyKSy7J1xH8b8CUgvrcVzOwqM1tiZkuam5szVpiIZIH5MLOkj4m1k7Jd6aiQ8fH4zexDwBbn3FIze//e1nPOLQQWAsyaNctlpjoRyQoX59aH30i62vy50zJQzOiXjSP+OcDZZrYO+DVwqpn9Mgt1iIjkpIwHv3Puy865GudcHXAh8Lhz7mOZrkNEJFfpPH4RkRyT1XvuOueeAJ7IZg0iIrlGR/wiIjlGwS8ikmMU/CIiOUbBLyKSYxT8IiI5RsEvIpJjFPwiIjkmq+fxi0ju6Yz62NgbZHskQHfUR3fMR9V5/8Xfl28kGPCRH/RTWZRHZVGIMQUhfD7LdsmjjoJfRNIq7qCxJ8SbnWHqe0J0Rv07X8vzxSnwx/EXlbOls49INE5PJEY8MSxj0G9MqihkamUhdZWFWXoHo4+CX0TSwl9UwTMtRazsCNMd8xOyOJMK+plQ2sWEcITKUJRAorN5/o2f40uJ0Tljcce2rn62dvaxobWHt7d28daWTnwGlR/6Akvrt3F87RjM9EngQCn4RQSAibWTaFzfkHS9mom1rG+o3+vra5o7+b8n11B99V281BpgckEfRxR3MLmgb2fQ74vfZ1QV51FVnMeRE0o41Tk2t/fxxqYOXuqdzXk/WEzfprdoX/xbut9cDAwyarvPD/FY8rYCQaKR/uRFjTIKfhEBoHF9w0GNib+1s4/bHn2Te19YT9BvdCx7iM9++ARKg8kDeF/MjPGlYcaXhvnT/A9y2f89xUsFQfLGL6CyKMR7Jlcwtapwt08A8+dO49aFP07+Xq664qBqG6kU/CJyUPqjcX78zFruXPQWPZEYH3tPLZ857TCqis+i9PzZQ9qWi/Qys6aUo6pLeHNTB8+v3cZfl29kQmmYkw+vYnxJeEjbG60U/CJywJbWb2fB/ct5Y3MHHzxyLF8+60imVhWlvV2fGUdMKOHwccWs2NjOP9a08JsX13Pk+GJOmlqZ9vZHOgW/iOy3nv4Y3/77Sn7xXD3jS8L86OOzOH36uIzX4fMZM6pLOWxcEUvWbeflhlZWb+mk9KQLicZJ6TuFXKTgF5H9Ehw7mQ/d8TRrmru47KQ6vnDGNIryshsleQE/cw6tZEZ1Kc+8tZXo+z7Gz9fHeF9FB4cW9qETgHanv4cikhLnHC81bGfCJbfS0Rvll1e8h6+dfVTWQ3+g0vwg82ZOYNM915Hni/O3zWX8fsMYmvuGT43DgfaGiCTV1RflkRWbqd/WTc/apTx4w1cpLwxlu6y96lv/GhfVbOP19nz+sa2IexvLOaqkhxPLOynwD3L6Z47REb+I7NParV386vkGGlt7+MC0Kprvv3FYh/4OPoOZpT1cWruVY0q7WdGez8/qK1m8rZC+2PDo+3EOemNGS7+frX0BmvsCtPT7016fjvhFZFDRWJxn32phWWMrlUUhzjuqmoqivGyXtd/CfscplZ3MLOlh8bYiXthexCttBRxX2o0vvyRjdcTjjjXNnby4bjsV8+bzi4YKOqI+Im7w4++wL05JMEZe9fQhr0XBLyLv0NLZx99f30RLZz/HTixjztQKAv6R3UFQHooxb3wbW/q6eG5bEc9tL6L6Uz/lP+57lUtPqmP6IUP7R6AvGuO1pjZeXLedJeu2saR+O63dEQDy646jLBhlUkGM4kCcwkAMAwyIOeiI+mmP+mmL+IlHeoa0LlDwi8gAzjmWN7Xx1OqthPw+zjnmkFE3ONrYvChnT2ilpd/PD/78In8syOc3S9Zz6NgizpoxnrlHjefICSX493NU0C0dvSxvbGNp/XaWrNvOssZW+qNxAKZUFXLG9PHMqhvD7LpyJlcVMT+FK4sBnt6ydr/fYzIKfhEBwJdfwl9e3cjbW7uYVFHA6UeOo3AYnbEz1CpCMbY9dCdr7r+VP7+ygb8t38T3Fr3Fdx9/i8KQn5k1pcysLmVCaT5VxXmUF4aIO0ckFqc3EmdDaw+N23to2NbN6xva2NzeB0AgcW3BpSdOYlZdObMmjRl2XWQZ/62a2UTg58A4vNGVFjrnbs90HSKyyzOrtzLhE3dQ39LNyYdVcuzEspwZ/bKsIMQlJ9ZxyYl1NHf08fTqZpatb+WV9a387B/19Mfie/3ZwpCfmjEFnDilghnV3h+Ko2vKyA/59/ozw0E2/pxHgc87514ys2JgqZk94pxbkYVaRHJaXzTGLQ+/ycKn3ibe18UFs4+jqnh4HZ1mUlVxHuceX8O5x9cAXtdXW0+E5o4+tnX14/cZQb+PUMDHhNIwpfnBEfkHMuPB75zbCGxMzHeY2UqgGlDwi2TQig3tzP/tMlZt6uBjJ9TyrY+eR9UFr2a7rH3yG+T3byMcbScY7yEY6yWQmAbjvQTivZiLcc3sIMd2PoURx5zDcMTNTww/cQsQS8z/8xEBeOsxCJdCXgnkFUO4BIIFYIaZUVYQoqxg+J++uj+y2oFnZnXAccDzg7x2FXAVQG1tbWYLExliqY51D8nHuz9Ysbhj4VNvc+sjb1BWEOKnn5jNB6aN5ZvRPm644YaUtjHUR7kleVDRvYaivi0U9W+huG8LRf2bKezfSn6kjfxoK+FIG9f+Vwm8eEbS7Z12Vj60PZB0vXkXFMAvz33H8mjcsbXbsbnTsanT0UmY8z7+b1A0DkomQFkdlNVCYSUjcTyIrAW/mRUBvwc+55xr3/N159xCYCHArFmzdKmdjGipjnUPex/vfig0tHTz+d8t48V12/l/M8bzzQ/P3O1irOsvPS15fc/eg3v8W0nXs1MX7PY8FO+hPLqFsmgzY6LNlEa3Mia6lbLoVq69rgRevnDnug6jO1hOZ6iKnmAZbeFqeoKl/OHen3PKOf9Cn6+AfgsRtRARyyPiCxJJPI/j4z/nX8vX//d2wEc8caKkjzg+F8VPDL+L4ncx/vfr/8myH34Son0Q6/em0T4CsT7GR7oZ39/FMf1drFvfBM//H8T6dn+TwULvD0BZLZRPgappMPZIqDoC8suS7qNsyUrwm1kQL/R/5Zy7Pxs1iOQS5xy/fnE9N/5lBT4zbj3/GD58XHV6+qfjMehu4V9mBnhv25+piGykMrqRkljrrnow2v1jaA1U8kbBcfzigcf4wKdvoTNvHB2hsXSFKon7gu/Y9NefvItbL35f0hJaehx9vuSnob6yOQ6lE5OuN/kLC3Dxduhrh7YmaK2H7fXQ2rBrft3TEOne9UNF42HsETD2KDjkWDjkeO+Pgy/710Nk46weA34MrHTO3Zrp9kVyzdqtXXz5/ld57u1tnDilgu+cfwzVZflDs3EXh+4W6NgIHZu8aecWcDF+dW4Bsc4n2RYYR1NoKq8EJ7AtMI7tgSraAxXEbFf83LL471TfMHdoakoXM++7gHApjBvkatp4HNrWQ/Mq77ElMV3yE4gmLsLKK4VDjuHbp+VxaM8rNIWm0OMvzuz7IDtH/HOAS4DlZrYssWyBc+5vWahFZNSKxOIsfOptbn9sNXkBH98+dyYXzJqIbz8vTBooHOvinGkBWPN4IuQ3e10kAP6Qd5RbPQuKx3HUZ+7hqht/SNyG96mNQ8bngzGTvMfhA76HiEVh6xvQ9BJseBk2vMT8E0OEtt0NQEtgLE2hqTTmTaUpbyqd/rK0l5qNs3qewbsyWUTSZGn9dr7ywHJWberg/80Yzw1nH8XYA7gtYX6sg5r+NVT3raGmbw1V0Y186sICaFoKRWNh3AwonuA9Cip2+6JzRXM8d0J/X/wBGHeU9zj+EgCKA8bdt3+Vmr41VPe9zbSelzm6ezEArf4KGvIOpyF8OA15h6elpNF7WZ5IDtrS3stNf1/F/S83Mb4kzMJL3sXco8an/PMFsXZq+tZQ07+Gmr63qIhuBiBiITaE6ng2/1iu++59PP2T68Gn+DhQ/THYFKpjU6iOJcWnYS5OZWTDzv0+reclju5eTBzjwcOHfj/rNycyCvRH4/z02bV897HVRGKOaz4wlX97/6HJh1xoa4L6Z2HdM6y6ppBpm673tmd5NIUms7JgFo2hQ9kcqiGe6JN/puE3Cv0h5sxHc6iG5lANLxedgs/FGN9fz6S+N3hxw++HvD399kRGMOccD762if956A3Wbu3ig0eO5asfms6kir2c0dLaAOuehfpnvOn2xABgeaW82RJn85QP05g3lS3Bapy6abImbn425E1hQ94UNnXeN+TbV/CLjFDPvd3Ct/++ilfWt3LY2KKdF2Lt5JwX7OueTRzVPwttiYvIwmUwaQ68+1+96fiZnO0PcOupH8jKe5HMUvCLjDArNrTznYff4PFVW5hQGuZ/PnI05x1fgx8HW1Z6IV+/GBoWQ3uT90MFFV7An/Rpbzp2+rA4n1yyQ8EvMkK81tTG7Y+t5pEVmykJB7jujMO4bHIb4aYH4DeJoO/Z7q1cNB4mneiFfN37vCtKR+DQApIeCn6RYW55Yxu3P/Ymj67cQnEQPjdlA58IPEjpP/4BTyauFC2fCkfMg9qTvMAfM1lBL3ul4BcZjnpaeXnZEr63uIXHthRRYt1cG/grl/keonRjn3fl6HGXeCFfexIUj8t2xTKCKPhFssgf72dMzzoqu9ZQ2f0WY7rWMO7aUzj/63fxgjuSUuDzxY9y6aE9lNS+G6qvhvFHQ6gg26XLCKbgF0m3/i6OqPQxedvTlPWup6yn0Zv2NlLSuxEfMXpdkN/HT2Zh7HzqQ1VUhyN89ZggF7z/NIrKLsj2O5BRZtQHf11tDfXrm5KuN2liNesaGkdN28Oe+d4xbO/e1ktFxva1cxCLQF8H9Gzj3DNOpq9jK+X5RkW+UVVoVBcb1cU+qkuMQ4p9lIWNldcUwcr5ALRFA7zdXcjS7kJW9E7n8eD7qC85nniwgGjnNrY/fjP1q57mHy7Olfuqxef3RsJMwh8IEo30H/h7zpBU7wUwpPbj32EqI5mOlP/Loz7469c3HdDY4SO97WEtHqc4GOeW228hFO8j5HoJxXsJuT5C8V78RPG7KD4X46EHfguLvu0NBBbr90I31g8u5oUw3q0a/uu4Fi5fMCfxdODtG9xuk3seWwa/u8wbVTKe2IaLJebjie0mXot0Q6QnMe3dNe92he39/wQwoNvFfBAq8h55Rd4dnULFXHzL3/jQJ79Mq7+SXl8hzf1BXm3PZ1VHPlFn1BX0cXzZNmrCET5/55Mpjd0/f+40bl344+TrXXVF0nWGg1TvBTCkXDzlfTia/i+P+uCXDOnv9kZrbG+C9o3Q1QzdW6Frqzdsb3dLYn4r9LbR/uUS2JT8CO+U08Lw5E3eEAH+UOIRHDBkgHcUdvqUgDcm+p52O0ozZh/ig02veQHt84P5vXV2zg9YXlAJwXzvNny7TfO9QM8v58xzL+LBWz7lLQvkQyBv0LNp7ln+J44L1rGmK49X2grY0BvCb45pRb0cX9ZFRSj5kbvIUFHwS3LxuBfq29d6l/y3NSUCfkPi0bjr/PGBfAEvPAsrvQuIJhzjzeeP4fNfuYEzz/84/ZZHvy9MxPLoS0yjFiSGd1/Uz19zDX39saQXG9Wa4R5PfrR1+KkLcG7Jge6Jd3hozflQcsg+19nYG6R0zr/wk/pKumN+SgNR3lfRwfTiHsJ+3VxOMk/BL55Ib+JOQutg21ov5HdMt9e/85ZzBRVQUg2lNTDx3V74ldZ40+JDvIAPl+71XPJbF3+Fmk+cmLSs/hgj8grTvrjx6JYSfts0hqe2FlM6ZxpVeRGOKWmnrqBfp9hLVin4c0n3tt0Dfds6L+i3r/WO3Af2jQcLoXwyVB7u3VRizGTvedkkL/CD+z+2ey5Y0R7mt03l/GFjGa2RABPC/Xxmyhb+40sL+NzNyfuIRTJBwT+axOPQseGdR+w7pr1tu69fNM4L9Mknw5i6XeE+ZrJ3xK7D0pTUd4f4y6ZS/ryxjFWd+YQsztxx7ZxfvY05FZ34DT7ftjnbZYrspOAfSZzz+tLbGr1Ha8O+u2R8Ae9G0uWTofpdu0K9fLIX9KHkN6OWwa3d2sUjKzYx/uO3csrT3l2S3lXWxQ1HNHHOhFbK9GWtDGMK/uGkvxs6N3lfnu4I97b1A+YbIdK1+8/srUtmzGQv9P36FQ+FSCzOi+u28fjKLTy+agtvb931e1hw+AbmjW+jOj+SxQpFUqdUSLdYBKK90N+VeHR608iu5298uhC+VQP9He/8+cIq70vTqsPh0NO8+Z2Pid7r6pIZcn3RGK82tvH82y08v3YbS+u3090fI+T3ccLUCi49qY5TjxhLbcWHuOoy9d3LyKLgT8Y5iPZ5V2r2d3jT3nbobfW+LO3ZnnjsmN+1vGtBMTxzy+Db9ed5XS2hQl7aGOfwsz7m3by6aByUVnuhXnKId364pFVvJMYbmzp4fUM7r29o4/UN7azc2E5fNA7AtHHFfORdNZw0tZL3HVaZ/HaGIsPc6P4X3N9FdbF5Fw7F+iDa701j/V6YD5jec24+/Oqj0NfphXtfe2LaAfEkH+F9QSgoh/wxkF/u9Z8fchx3PvRjvnjxXAiEvUG1gkWQV+h1z/iDO3/8ovkLuPC+m9K7L3JUX9zY1h+gpT9Ac1+Awhmncueit6hv6WJdSzcNLd1sau/duX5xOMBRh5RwyQmTmD25nHfXlTOmMJTFdyAy9EZ38D/0FRrnF8OSu/a+ji8A/jzedYgPOrd4V2SWTfSmOx6hIsgrGbCsaFfI54/xjtwH6W750oe/zxe/nPxc9ZEkEovT3hOhLfFo743umu+J0N0fpTcSpzcS86bRGH2RGH3ROH2RODHniDvH+Etu4d7GcnDeSaR7u4xpwifu4Mzbnkpa14RP3MGZz44n4ozumI+uqI/umJ+o2/33UjlvPjc/9AZVxXlMKi/gpEMrmFReyOHjijjqkFImluenNCaLyEiWleA3szOB2wE/cJdzLj2HuzM/ypXXf5+7/uNC71L/QN6AaR4EQjsHAZt26gKcezItZQwn8bijsz9Ke0+E9p4o7b1eYLf3RncGenvvriBv7xkQ7L0Ruvv3fbaK32eEAz7CQT/hoJ+8gI+8oJ9w0EdewEfI58cM4r0dFPi9rhQDDDfoVxUNrZuoLT8h6ft6qXUTtQXlBM1REIhT6I+T7/emY0JRKhOPE664ibbN6ykIje5jHpF9yfi/fjPzA3cCpwONwItm9ifn3Iohb6xuDj9+OcJdY48cks055xLTXUeozrkB87DjmXOAP0hvzPZ43Uu3gUe5FipgW1c/0VicSNx505gjGo8TjTkisTjReGKaWN4bidPdH6OnP0pPJObNR2L09HuP7sR8d390t4Dv6IvikowSUJwXoCQ/SEl+kNL8AJMqCijND1K6c1mQkvzArmXhXa+Fg/6U9qVdeQLXnZ58cKzHH/gmC++/Mel6P7p0Ngs/k/xL1mjrJoW+5Lxs/A94N/CWc+5tADP7NXAOMOTB/42/rGDitfdx5KOhnWG3W7fCgACv/cIfmLrgb97iPcL8QE36wgMc8Wjy9Wqv/S3Hf+ORA28oIRTwURDykx/0k5+YFoT8TCgNM218cSKkE6Ee9sLbm3rPi8MBisMBAv6RN0SCiKTO3MEk24E0aPYR4Ezn3JWJ55cA73HOfXqP9a4Crko8nQYkH6c2MyqBrdkuIgnVePCGe30w/Gsc7vXB6K9xknOuas+Fw/Yzr3NuIbAw23XsycyWOOdmZbuOfVGNB2+41wfDv8bhXh/kbo3Z+EzfBEwc8LwmsUxERDIgG8H/InCYmU02sxBwIfCnLNQhIpKTMt7V45yLmtmngYfwTuf8iXPu9UzXcRCGXffTIFTjwRvu9cHwr3G41wc5WmPGv9wVEZHs0nl7IiI5RsEvIpJjFPx7YWZnmtkbZvaWmV03yOtXm9lyM1tmZs+Y2fThVuOA9c4zM2dmGT1tLYV9eJmZNSf24TIzuzKT9aVSY2Kd881shZm9bmb3DKf6zOx/B+y/N82sNZP1pVhjrZktMrOXzexVMztrGNY4ycweS9T3hJnVZLi+n5jZFjN7bS+vm5l9N1H/q2Z2/EE16JzTY48H3pfOa4ApQAh4BZi+xzolA+bPBh4cbjUm1isGngKeA2YNp/qAy4DvDfPf82HAy8CYxPOxw6m+Pdb/DN7JEsNtHy4EPpWYnw6sG4Y1/g64NDF/KvCLDNd4MnA88NpeXj8L+Dve0FYnAM8fTHs64h/czmElnHP9wI5hJXZyzrUPeFrI3geYTJekNSZ8A/hvoHeQ19Ip1fqyKZUa/xW40zm3HcA5t2WY1TfQRcC9Galsl1RqdEBJYr4U2JDB+iC1GqcDjyfmFw3yelo5554Ctu1jlXOAnzvPc0CZmU040PYU/IOrBtYPeN6YWLYbM7vGzNYA/wN8NkO17ZC0xsTHwYnOub9msrCElPYhcF7io+t9ZjZxkNfTKZUaDwcON7Nnzey5xMiymZLqPsTMJgGT2RVemZJKjV8DPmZmjcDf8D6ZZFIqNb4CnJuY/zBQbGYVGagtVSn/W0iFgv8gOOfudM5NBf4D+M9s1zOQmfmAW4HPZ7uWffgzUOecOxp4BPhZlusZTACvu+f9eEfUPzKzsmwWtBcXAvc554bjXd4vAu52ztXgdVn8IvHvczj5AnCKmb0MnII3msBw3JdDYrjt/OFif4eV+DXwz+ksaBDJaiwGZgBPmNk6vH7BP2XwC96k+9A51+Kc60s8vQt4V4Zq2yGV33Mj8CfnXMQ5txZ4E+8PwXCpb4cLyXw3D6RW4xXAbwGcc4uBMN7AY5mSyr/FDc65c51zxwFfSSxrzViFyQ3tUDeZ/AJjpDzwjvLexvvovOPLoKP2WOewAfP/BCwZbjXusf4TZPbL3VT24YQB8x8Gnhtu+xA4E/hZYr4S7+N2xXCpL7HeEcA6EhdkDsN9+HfgssT8kXh9/BmrNcUaKwFfYv6bwNezsC/r2PuXu/PY/cvdFw6qrUy/uZHywPtI+ibe2QBfSSz7OnB2Yv524HVgGd6XQXsN3WzVuMe6GQ3+FPfhtxP78JXEPjxiuO3DxH+0W/HuF7EcuHA41Zd4/jXgpkzvu/3Yh9OBZxO/52XA3GFY40eA1Yl17gLyMlzfvcBGIIL3KfMK4Grg6gH/Du9M1L/8YP8va8gGEZEcoz5+EZEco+AXEckxCn4RkRyj4BcRyTEKfhGRHKPgl5xlZneb2UcGWb7OzFK+wMjM3m9mf0lhvZsTI3zevL+1igyljN96USSHXQWUuzQNq2BmAedcNB3bltFFR/wyIplZoZn91cxeMbPXzOwCM3uXmT1pZkvN7KEdoxea2VQzezCx/GkzO2KQ7X0j8QnAn1j0GTN7KXHPhSMGtPkTM3shMbb8O0ZwNLOvmdkvzGyxma02s39NLP8TUAQsTdRaZWa/N7MXE485ZuZLfNooG7C91WY2brD192jvWeAXQ7ybZZTSEb+MVGcCG5xz8wDMrBTvkvZznHPNZnYB3qX3l+ONB3+1c261mb0H+D7emOskfvZmvLGNPuGcc2YGsNU5d7yZ/RveAF5X4o3h8rhz7vJEOL9gZo8OUtvReJfVFwIvm9lfnXNnm1mnc+7YRJv3AP/rnHvGzGqBh5xzR5rZH/GGr/hpotZ659zmwdbHG/4AvCtj3+uc6zn43Sq5QMEvI9Vy4BYz+2/gL8B2vEHpHkkEtx/YaGZFwEnA7xLLAfIGbOereDe1uGqP7d+fmC5l13C9c4GzzewLiedhoHaQ2v6YCOEeM1uENx78H/ZY54PA9AE1lSRq/Q3wX8BP8QZe+02S9cEbRE6hLylT8MuI5Jx7M3G/gbOAG/HGoX/dOXfiwPXMrARo3XGkPYgXgXeZWblzbuCNMHaMGhpj1/8TA85zzr2xRxvj9iwvyXPwullPcM7tdoMcM1sMHGpmVXgjvt6YZH2ArsHfmsjg1McvI5KZHQJ0O+d+CdwMvAeoMrMTE68Hzewo590pba2ZfTSx3MzsmAGbehC4CfirmRUnafYhvL5/S2zruL2sd46ZhRM38ng/3h+XPT3MgBuSmNmxAM4bPOsBvIHhVjrnWva1vsiBUPDLSDUTr499GXA9XvfIR4D/NrMdo0CelFj3YuCKxPLXeedtNH8H/AjvfgX5+2jzG0AQeNXMXk88H8yreKONPgd8wzk32K0GPwvMMu/uYyvwRmLc4TfAx9jVzZNsfZH9otE5RYaQmX0N6HTOfSfbtYjsjY74RURyjI74RURyjI74RURyjIJfRCTHKPhFRHKMgl9EJMco+EVEcsz/Bx1qnaBRGV0EAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "daa = dda[(dda.week >= 38) & dda.seekhelpfever.notna()].groupby(['stratumid', 'treatment']).agg({'seekhelpfever': 'mean', 'userid': 'count'}).reset_index()\n",
    "sns.histplot(da, x='seekhelpfever', hue=\"treatment\", bins=30, kde=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddmerge = da.rename(columns = {'seekhelpfever': 'pre_seekhelpfever'}).merge(daa[['stratumid', 'seekhelpfever']], on=['stratumid'])\n",
    "\n",
    "ddmerge['delta'] = ddmerge.seekhelpfever - ddmerge.pre_seekhelpfever\n",
    "\n",
    "smf.ols('delta ~ treatment', ddmerge).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 886,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>OLS Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>      <td>seekhelpfever</td>  <th>  R-squared:         </th> <td>   0.001</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                   <td>OLS</td>       <th>  Adj. R-squared:    </th> <td>  -0.012</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>             <td>Least Squares</td>  <th>  F-statistic:       </th> <td> 0.05265</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>             <td>Tue, 13 Apr 2021</td> <th>  Prob (F-statistic):</th>  <td> 0.819</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                 <td>13:59:00</td>     <th>  Log-Likelihood:    </th> <td>  58.499</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Observations:</th>      <td>    80</td>      <th>  AIC:               </th> <td>  -113.0</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Residuals:</th>          <td>    78</td>      <th>  BIC:               </th> <td>  -108.2</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Model:</th>              <td>     1</td>      <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>nonrobust</td>    <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>t</th>      <th>P>|t|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    0.8742</td> <td>    0.019</td> <td>   46.875</td> <td> 0.000</td> <td>    0.837</td> <td>    0.911</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.0061</td> <td>    0.026</td> <td>    0.229</td> <td> 0.819</td> <td>   -0.046</td> <td>    0.059</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "  <th>Omnibus:</th>       <td> 9.253</td> <th>  Durbin-Watson:     </th> <td>   1.496</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Prob(Omnibus):</th> <td> 0.010</td> <th>  Jarque-Bera (JB):  </th> <td>   9.759</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Skew:</th>          <td>-0.855</td> <th>  Prob(JB):          </th> <td> 0.00760</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Kurtosis:</th>      <td> 3.030</td> <th>  Cond. No.          </th> <td>    2.62</td>\n",
       "</tr>\n",
       "</table><br/><br/>Notes:<br/>[1] Standard Errors assume that the covariance matrix of the errors is correctly specified."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                            OLS Regression Results                            \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   R-squared:                       0.001\n",
       "Model:                            OLS   Adj. R-squared:                 -0.012\n",
       "Method:                 Least Squares   F-statistic:                   0.05265\n",
       "Date:                Tue, 13 Apr 2021   Prob (F-statistic):              0.819\n",
       "Time:                        13:59:00   Log-Likelihood:                 58.499\n",
       "No. Observations:                  80   AIC:                            -113.0\n",
       "Df Residuals:                      78   BIC:                            -108.2\n",
       "Df Model:                           1                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      0.8742      0.019     46.875      0.000       0.837       0.911\n",
       "treatment      0.0061      0.026      0.229      0.819      -0.046       0.059\n",
       "==============================================================================\n",
       "Omnibus:                        9.253   Durbin-Watson:                   1.496\n",
       "Prob(Omnibus):                  0.010   Jarque-Bera (JB):                9.759\n",
       "Skew:                          -0.855   Prob(JB):                      0.00760\n",
       "Kurtosis:                       3.030   Cond. No.                         2.62\n",
       "==============================================================================\n",
       "\n",
       "Notes:\n",
       "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n",
       "\"\"\""
      ]
     },
     "execution_count": 886,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smf.ols('seekhelpfever ~ treatment', da).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sns.histplot(d, x='seekhelpfever', hue=\"treatment\", bins=30, kde=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 679,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "permutations = pd.read_csv('outs/cluster-assignments-2.csv').values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 702,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "@curry\n",
    "def calc_te(df, t):\n",
    "    vals = df.assign(treatment = t).groupby('treatment').mean().reset_index().sort_values('treatment').seekhelpfever.values\n",
    "    return vals[1] - vals[0]\n",
    "\n",
    "\n",
    "tes = np.array([calc_te(d)(permutations[:, i]) for i in range(1000)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 703,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x7f206a02c940>"
      ]
     },
     "execution_count": 703,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.displot(tes, bins=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 704,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.042"
      ]
     },
     "execution_count": 704,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "((tes >= .0249).sum() + (tes <= -0.0249).sum()) / tes.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 714,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x7f20702aec40>"
      ]
     },
     "execution_count": 714,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.displot(ddb[ddb.seekhelpfever.notna()].groupby('stratumid').count().userid.sort_values(), bins=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "@curry\n",
    "def calc_te(df, ta, t):\n",
    "    vals = df.drop(columns=['treatment']) \\\n",
    "             .merge(ta.assign(treatment = t), left_on='stratumid', right_on='disthash') \\\n",
    "             .groupby('treatment') \\\n",
    "             .mean() \\\n",
    "             .reset_index() \\\n",
    "             .sort_values('treatment') \\\n",
    "             .seekhelpfever \\\n",
    "             .values\n",
    "\n",
    "    return vals[1] - vals[0]\n",
    "\n",
    "\n",
    "tes = np.array([calc_te(ddb[(ddb.wave == '0') & ddb.kutcha & ddb.dwelling.notna()], treatment_assignment)(permutations[:, i]) for i in range(1000)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sns.displot(tes, bins=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "te_group = calc_te(ddb, treatment_assignment, treatment_assignment.treatment.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "te_group"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "((tes > te_group).sum() + (tes < -te_group).sum()) / tes.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 644,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  seekhelpfever\n",
       "0          1.0              0.825905\n",
       "           0.0              0.174095\n",
       "1          1.0              0.791606\n",
       "           0.0              0.208394\n",
       "Name: seekhelpfever, dtype: float64"
      ]
     },
     "execution_count": 644,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dda.groupby('treatment').seekhelpfever.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 645,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  seekhelpfever\n",
       "0          1.0              0.938482\n",
       "           0.0              0.061518\n",
       "1          1.0              0.958275\n",
       "           0.0              0.041725\n",
       "Name: seekhelpfever, dtype: float64"
      ]
     },
     "execution_count": 645,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb.groupby('treatment').seekhelpfever.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "vv = v.groupby('pair_id').head(1).reset_index(drop=True)\n",
    "\n",
    "smf.ols('delta ~ 1', vv).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 308,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.02590120769547678"
      ]
     },
     "execution_count": 308,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.mean(vv.delta)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "np.sqrt(np.var(vv.delta)) / np.sqrt(vv.shape[0] - 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 280,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.2301406717356983"
      ]
     },
     "execution_count": 280,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v = d[d.pucca == False].groupby(['stratumid', 'treatment']).agg({'seekhelpfever': 'mean', 'userid': 'count'}).reset_index()\n",
    "\n",
    "np.sqrt(v[v.treatment == 1].seekhelpfever.var() + v[v.treatment == 0].seekhelpfever.var())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import scipy\n",
    "\n",
    "scipy.stats.t.pdf(13.4, 7, 0, 13.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 306,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "v = d[d.pucca == False] \\\n",
    "    .groupby(['stratumid', 'treatment']) \\\n",
    "    .agg({'seekhelpfever': 'mean', 'userid': 'count'}) \\\n",
    "    .rename(columns={'userid': 'count'}) \\\n",
    "    .pipe(lambda df: df.assign(rank = center(rankdata(df.seekhelpfever)))) \\\n",
    "    .reset_index() \\\n",
    "    .merge(pair_lookup) \\\n",
    "    .groupby('pair_id') \\\n",
    "    .apply(lambda df: dif(df, 'seekhelpfever')) \\\n",
    "    .reset_index(drop=True) \\\n",
    "    .sort_values('count')\n",
    "\n",
    "# np.mean(v.delta)\n",
    "\n",
    "# sns.distplot(v.delta, bins=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# build a sense of statistical uncertainty per cluster\n",
    "# increase repr in that cluster\n",
    "with pd.option_context(\"display.max_rows\", 160, \"display.max_columns\", 10):\n",
    "    display(v)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "d[d.pucca == False] \\\n",
    "    .groupby(['stratumid', 'treatment']) \\\n",
    "    .agg({'seekhelpfever': 'mean', 'userid': 'count'}) \\\n",
    "    .rename(columns={'userid': 'count'}) \\\n",
    "    .reset_index() \\\n",
    "    .merge(pair_lookup) \\\n",
    "    .sort_values('seekhelpfever') \\\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.002882574759349029"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "v = d.groupby(['stratumid', 'treatment']).malaria4months.mean().reset_index().merge(pair_lookup).groupby('pair_id').apply(lambda df: dif(df, 'malaria4months')).delta.values\n",
    "\n",
    "np.mean(v)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 923,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def aggs(df, sums, anys, means):\n",
    "    r = df.iloc[0]\n",
    "    # fil = df.wave > 1\n",
    "    fil = df.week < 36 # 36 = first week of september\n",
    "    for var in sums:\n",
    "        r[var] = df[(fil)][var].sum()\n",
    "    for var in anys:\n",
    "        r[var] = int(df[(fil)][var].sum() > 0)\n",
    "    for var in means:\n",
    "        r[var] = df[(fil)][var].mean()\n",
    "\n",
    "    r['wave'] = None\n",
    "    r['waves_answered'] = df[(fil)].shape[0]\n",
    "    return r\n",
    "\n",
    "\n",
    "# feature matrix\n",
    "\n",
    "# user, sum bednet sleeps in follows, if malaria at any point, binary controls (malaria in last 5 years?)\n",
    "\n",
    "mat = dda \\\n",
    "    .sort_values('week') \\\n",
    "    .groupby('userid') \\\n",
    "    .apply(lambda df: aggs(df, \n",
    "                           sums=['fever2weeks', 'seekhelpfever'], \n",
    "                           anys=['malaria2weeks'], \n",
    "                           means=['sleepundernet', 'longsleeves', 'bednet_share'])) \\\n",
    "    .reset_index(drop=True) \\\n",
    "    .dropna(subset=['sleepundernet'])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 926,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "mat['seekhelp_if_fever'] = mat.seekhelpfever / mat.fever2weeks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 921,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  seekhelp_if_fever\n",
       "0          1.000000             0.880184\n",
       "           0.000000             0.110599\n",
       "           0.500000             0.009217\n",
       "1          1.000000             0.862944\n",
       "           0.000000             0.111675\n",
       "           0.500000             0.010152\n",
       "           0.333333             0.005076\n",
       "           0.666667             0.005076\n",
       "           0.750000             0.005076\n",
       "Name: seekhelp_if_fever, dtype: float64"
      ]
     },
     "execution_count": 921,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mat.groupby('treatment')['seekhelp_if_fever'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 927,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  seekhelp_if_fever\n",
       "0          1.0                  0.855556\n",
       "           0.0                  0.136111\n",
       "           0.5                  0.008333\n",
       "1          1.0                  0.786280\n",
       "           0.0                  0.200528\n",
       "           0.5                  0.013193\n",
       "Name: seekhelp_if_fever, dtype: float64"
      ]
     },
     "execution_count": 927,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mat.groupby('treatment')['seekhelp_if_fever'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 922,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>OLS Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>    <td>seekhelp_if_fever</td> <th>  R-squared:         </th> <td>   0.000</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                   <td>OLS</td>        <th>  Adj. R-squared:    </th> <td>  -0.002</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>             <td>Least Squares</td>   <th>  F-statistic:       </th> <td> 0.06341</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>             <td>Tue, 13 Apr 2021</td>  <th>  Prob (F-statistic):</th>  <td> 0.801</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                 <td>14:11:02</td>      <th>  Log-Likelihood:    </th> <td> -112.56</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Observations:</th>      <td>   414</td>       <th>  AIC:               </th> <td>   229.1</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Residuals:</th>          <td>   412</td>       <th>  BIC:               </th> <td>   237.2</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Model:</th>              <td>     1</td>       <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>nonrobust</td>     <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>t</th>      <th>P>|t|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    0.8848</td> <td>    0.022</td> <td>   40.943</td> <td> 0.000</td> <td>    0.842</td> <td>    0.927</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>   -0.0079</td> <td>    0.031</td> <td>   -0.252</td> <td> 0.801</td> <td>   -0.069</td> <td>    0.054</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "  <th>Omnibus:</th>       <td>195.942</td> <th>  Durbin-Watson:     </th> <td>   2.089</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Prob(Omnibus):</th> <td> 0.000</td>  <th>  Jarque-Bera (JB):  </th> <td> 611.742</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Skew:</th>          <td>-2.356</td>  <th>  Prob(JB):          </th> <td>1.45e-133</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Kurtosis:</th>      <td> 6.641</td>  <th>  Cond. No.          </th> <td>    2.57</td> \n",
       "</tr>\n",
       "</table><br/><br/>Notes:<br/>[1] Standard Errors assume that the covariance matrix of the errors is correctly specified."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                            OLS Regression Results                            \n",
       "==============================================================================\n",
       "Dep. Variable:      seekhelp_if_fever   R-squared:                       0.000\n",
       "Model:                            OLS   Adj. R-squared:                 -0.002\n",
       "Method:                 Least Squares   F-statistic:                   0.06341\n",
       "Date:                Tue, 13 Apr 2021   Prob (F-statistic):              0.801\n",
       "Time:                        14:11:02   Log-Likelihood:                -112.56\n",
       "No. Observations:                 414   AIC:                             229.1\n",
       "Df Residuals:                     412   BIC:                             237.2\n",
       "Df Model:                           1                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      0.8848      0.022     40.943      0.000       0.842       0.927\n",
       "treatment     -0.0079      0.031     -0.252      0.801      -0.069       0.054\n",
       "==============================================================================\n",
       "Omnibus:                      195.942   Durbin-Watson:                   2.089\n",
       "Prob(Omnibus):                  0.000   Jarque-Bera (JB):              611.742\n",
       "Skew:                          -2.356   Prob(JB):                    1.45e-133\n",
       "Kurtosis:                       6.641   Cond. No.                         2.57\n",
       "==============================================================================\n",
       "\n",
       "Notes:\n",
       "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n",
       "\"\"\""
      ]
     },
     "execution_count": 922,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smf.ols('seekhelp_if_fever ~ treatment', mat).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1130,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>OLS Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>    <td>seekhelp_if_fever</td> <th>  R-squared:         </th> <td>   0.008</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                   <td>OLS</td>        <th>  Adj. R-squared:    </th> <td>   0.007</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>             <td>Least Squares</td>   <th>  F-statistic:       </th> <td>   5.866</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>             <td>Wed, 14 Apr 2021</td>  <th>  Prob (F-statistic):</th>  <td>0.0157</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                 <td>08:17:56</td>      <th>  Log-Likelihood:    </th> <td> -322.82</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Observations:</th>      <td>   739</td>       <th>  AIC:               </th> <td>   649.6</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Residuals:</th>          <td>   737</td>       <th>  BIC:               </th> <td>   658.8</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Model:</th>              <td>     1</td>       <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>nonrobust</td>     <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>t</th>      <th>P>|t|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    0.8597</td> <td>    0.020</td> <td>   43.496</td> <td> 0.000</td> <td>    0.821</td> <td>    0.899</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>   -0.0668</td> <td>    0.028</td> <td>   -2.422</td> <td> 0.016</td> <td>   -0.121</td> <td>   -0.013</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "  <th>Omnibus:</th>       <td>199.896</td> <th>  Durbin-Watson:     </th> <td>   2.087</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Prob(Omnibus):</th> <td> 0.000</td>  <th>  Jarque-Bera (JB):  </th> <td> 381.981</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Skew:</th>          <td>-1.695</td>  <th>  Prob(JB):          </th> <td>1.13e-83</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Kurtosis:</th>      <td> 3.959</td>  <th>  Cond. No.          </th> <td>    2.65</td>\n",
       "</tr>\n",
       "</table><br/><br/>Notes:<br/>[1] Standard Errors assume that the covariance matrix of the errors is correctly specified."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                            OLS Regression Results                            \n",
       "==============================================================================\n",
       "Dep. Variable:      seekhelp_if_fever   R-squared:                       0.008\n",
       "Model:                            OLS   Adj. R-squared:                  0.007\n",
       "Method:                 Least Squares   F-statistic:                     5.866\n",
       "Date:                Wed, 14 Apr 2021   Prob (F-statistic):             0.0157\n",
       "Time:                        08:17:56   Log-Likelihood:                -322.82\n",
       "No. Observations:                 739   AIC:                             649.6\n",
       "Df Residuals:                     737   BIC:                             658.8\n",
       "Df Model:                           1                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      0.8597      0.020     43.496      0.000       0.821       0.899\n",
       "treatment     -0.0668      0.028     -2.422      0.016      -0.121      -0.013\n",
       "==============================================================================\n",
       "Omnibus:                      199.896   Durbin-Watson:                   2.087\n",
       "Prob(Omnibus):                  0.000   Jarque-Bera (JB):              381.981\n",
       "Skew:                          -1.695   Prob(JB):                     1.13e-83\n",
       "Kurtosis:                       3.959   Cond. No.                         2.65\n",
       "==============================================================================\n",
       "\n",
       "Notes:\n",
       "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n",
       "\"\"\""
      ]
     },
     "execution_count": 1130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smf.ols('seekhelp_if_fever ~ treatment', mat).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 913,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def as_type(series, type_):\n",
    "    return series.astype(type_)\n",
    "\n",
    "def quantile(series, q):\n",
    "    return series > series.quantile(q)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dd['unet'] = dd.under_net.astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dd['malaria_now_int'] = dd.malaria_now.astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dd[dd.wave == 0].malaria_now"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 412,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1     7306\n",
       "7     5538\n",
       "8     5066\n",
       "9     4568\n",
       "10    2959\n",
       "3     2949\n",
       "6      568\n",
       "5      560\n",
       "11     328\n",
       "4      127\n",
       "2       99\n",
       "Name: week, dtype: int64"
      ]
     },
     "execution_count": 412,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb.week.value_counts("
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 479,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>      <th>Type</th> <th>Post. Mean</th> <th>Post. SD</th>  <th>SD</th>   <th>SD (LB)</th> <th>SD (UB)</th>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th>   <td>M</td>    <td>2.6096</td>    <td>0.1183</td>    <td></td>       <td></td>        <td></td>    \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th>   <td>M</td>    <td>0.5308</td>    <td>0.1901</td>    <td></td>       <td></td>        <td></td>    \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>a</th>           <td>V</td>    <td>-0.3955</td>   <td>0.0790</td>  <td>0.673</td>  <td>0.575</td>   <td>0.789</td> \n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary2.Summary'>\n",
       "\"\"\"\n",
       "               Binomial Mixed GLM Results\n",
       "========================================================\n",
       "          Type Post. Mean Post. SD   SD  SD (LB) SD (UB)\n",
       "--------------------------------------------------------\n",
       "Intercept    M     2.6096   0.1183                      \n",
       "treatment    M     0.5308   0.1901                      \n",
       "a            V    -0.3955   0.0790 0.673   0.575   0.789\n",
       "========================================================\n",
       "Parameter types are mean structure (M) and variance\n",
       "structure (V)\n",
       "Variance parameters are modeled as log standard\n",
       "deviations\n",
       "\"\"\""
      ]
     },
     "execution_count": 479,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m = sm.BinomialBayesMixedGLM.from_formula('seekhelpfever ~ treatment', {'a': '0 + C(stratumid)'}, ddb[(ddb.week < 20) & ddb.dwelling.notna() & ddb.seekhelpfever.notna()])\n",
    "mm = m.fit_vb()\n",
    "\n",
    "mm.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 481,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.9308615796566533, 0.9581133929422073)"
      ]
     },
     "execution_count": 481,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(2.6), expit(2.6 + .53)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 436,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[33mWARNING: You are using pip version 20.2.3; however, version 21.0.1 is available.\r\n",
      "You should consider upgrading via the '/usr/local/bin/python -m pip install --upgrade pip' command.\u001b[0m\r\n"
     ]
    }
   ],
   "source": [
    "! pip install --quiet bambi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 587,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0    0.641818\n",
       "1.0    0.358182\n",
       "Name: fever4months, dtype: float64"
      ]
     },
     "execution_count": 587,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[(ddb.malaria4months == 1)].fever4months.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 588,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Yes           0.559859\n",
       "No            0.352113\n",
       "Don’t know    0.088028\n",
       "Name: testmalaria, dtype: float64"
      ]
     },
     "execution_count": 588,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[(ddb.malaria4months == 1)].testmalaria.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 718,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "d = ddb[ddb.seekhelpfever.notna() & ddb.dwelling.notna()].reset_index(drop=True)\n",
    "d.loc[d.seekhelpfever == 'No', 'timeseekhelpfever'] = 'Did not seek help'\n",
    "d = d.dropna(subset=['timeseekhelpfever']).reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 719,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from pandas.api.types import CategoricalDtype\n",
    " \n",
    "cat_type = CategoricalDtype(categories=['Less than 24 hours', '24-48 hours', 'More than 2 days', 'Did not seek help'], ordered=True)\n",
    "\n",
    "d['timeseekhelpfever'] = d.timeseekhelpfever.astype(cat_type)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import theano.tensor as tt\n",
    "import pymc3 as pm\n",
    "\n",
    "stratumid = d.stratumid.astype('category').cat.codes\n",
    "treatment = d.treatment\n",
    "P = 80\n",
    "y = d.timeseekhelpfever.cat.codes\n",
    "levels = 4\n",
    "\n",
    "with pm.Model() as ordinal_model:\n",
    "    intercept = pm.Normal('intercept', mu=0, sigma=2)\n",
    "    \n",
    "    stratumid_sigma = pm.HalfCauchy(\"stratumid_sigma\", 2.0)\n",
    "     \n",
    "    a_offset = pm.Normal(\"a_offset\", mu=0, sigma=1, shape=P)\n",
    "    a = pm.Deterministic(\"a\", a_offset*stratumid_sigma)\n",
    "    b = pm.Normal(\"b\", mu=0, sigma=2)\n",
    "\n",
    "    eta = intercept + a[stratumid] + b*treatment    \n",
    "    cutpoints = pm.Normal(\"cutpoints\", mu=np.array([-1,0,1]), sd=1, shape=3, transform=pm.distributions.transforms.ordered)\n",
    "    yl = pm.OrderedLogistic('obs', cutpoints=cutpoints, eta=eta, observed=y)\n",
    "    trace = pm.sample(draws=2_000, tune=2_000, return_inferencedata=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 779,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:title={'center':'b'}>"
      ]
     },
     "execution_count": 779,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbgAAAEoCAYAAAAqrOTwAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Z1A+gAAAACXBIWXMAAAsTAAALEwEAmpwYAABBDklEQVR4nO3dd3wUZf7A8c+T3htJSCMBAiF0QhOUIliAsx3Y0DuV4xTx1FPP0/N3nBTRE+thO8+Ts5+IYjkBBSugFOk1hFATSCOQ3pPd+f0xu8mmt0022f2+X699ZTPzzMx3Jpv57vPM88woTdMQQggh7I2TrQMQQgghOoIkOCGEEHZJEpwQQgi7JAlOCCGEXZIEJ4QQwi5JghNCCGGXJMEJ0QmUUouVUppSysXWsQjhKCTBCSGEsEuS4IQQQtglSXBCdK6BSqkflVIlSqkMpdQTSin5PxSiA8g/lhCd6wvgO+DXwIfA48BCG8YjhN2SC95CdK43NU1bZnr/jVLKD3hYKbVc07Q8G8YlhN2RGpwQnevjOr9/BPgAQ2wQixB2TRKcEJ0rq5HfIzs7ECHsnSQ4ITpXz0Z+T+vsQISwd5LghOhcN9X5fTZQBBy0QSxC2DXpZCJE57rLNCxgJzANuBNYrGlavm3DEsL+SA1OiM51HXAF8CXwW+BJYKlNIxLCTilN02wdgxBCCGF1UoMTQghhlyTBCSGEsEuS4IQQQtglSXBCCCHskiQ4IYQQdqml4+Ckq6UQQoiuSDU2Q2pwQggh7JIkOCGEEHZJEpwQQgi7JAlOCCGEXZIEJ4QQwi5JghNCCGGXJMEJIYSwS5LghBBC2CVJcEIIIeySPNFbiA6w70wez21IIq+kklsviubWsdEo1egNF4QQHUBqcEJYWXJWIbe+uZ3j54pwcVIs+PwQz39z1NZhCeFwpAYnhBUZjRr/99lB3F2cWHPfBIJ93Pnr5wd57ccTJPQK5PJBPW0dohAOQ2pwQljRj0fPsTsll8dmxBPq54GTk2LJdYOJD/Nl0ZeHKas02DpEIRyGJDghrOj97SmE+Loza2RU9TR3F2cWXTOYtLxS3t162nbBCeFgJMEJYSVpeaVsSs7mlrHRuDrX/tcaH9uDCf2CWfHzKanFCdFJJMEJYSXrD2WiaTAzIbLB+fMnx5JdWM6X+9M7OTIhHJMkOCGsZMOhTOLDfOkT7N3g/Ev69SA2xJuPdqR2cmRCOCZJcEJYwYWicnam5DBtcFijZZRSzB4TzZ7UPJKzCjsxOiEckyQ4Iaxg64kLaBpMiQ9tstzMkZE4Oym+3CfNlEJ0NElwQljBz8fO4+fhwtBI/ybLBfu4M65vEF8dzEDTtE6KTgjHJAlOiHbSNI2fj5/n4thgnJ2avx3Xr4aGc/J8MUmZ0kwpREeSBCdEO6VcKCEtr5RL+ge3qPyVg/TrdD8knevIsIRweJLghGinbScvAHBJbI8WlQ/xdWdQuB+bk7M7MiwhHJ4kOCHaaXdKLkHebo0OD2jI5AEh7E7JpbCssgMjE8KxSYITop32pOYyMjqgVY/DmdQ/hCqjxrYTFzowMiEcmyQ4Idoht7iCk9nFjIwJbNVyo2IC8XJz5ufj5zsoMiGEJDgh2mHvmVwARka3LsG5uTgxMjqQnadzOyIsIQSS4IRolz0peTg7KYZFNT3+rSGjYgI5mlkg1+GE6CCS4IRohz2puQwM98XLrfXPDh7dOxCjBntT86wfmBBCEpwQbVVlMLLvTF6rmyfNEqIDcVKwK0WaKYXoCJLghGijo1mFlFQYGNXKDiZmPu4uxIf5sTslx8qRCSFAEpwQbbbH1LTY1hoc6Nfh9qbmUWUwWikqIYSZJDgh2mhvSi7BPu5EBXq2eR2jewdSUmGQ+1IK0QEkwQnRRrvbMMC7ruFRAQAcSsu3UlRCCDNJcEK0wfmiclIulLT5+ptZTA8vfD1cOGBKcFu2bOGiiy7Cw8ODPn368PLLL7doPWlpacycORNfX1+Cg4O57777KCkpqZ5fUFDAokWLGDt2LP7+/oSFhTFz5kySk5NrraeiooJHHnmEiRMn4unp2a7kLYStSYITog3MXftbeweTupRSDI305+DZfI4fP860adPo06cPX331FXfffTd/+tOfWLFiRZPrqKysZNq0aaSkpPDRRx/x0ksv8cknnzBv3rzqMqmpqbz55ptMmzaN1atX88Ybb5CRkcFFF13EmTNnqsuVlJSwYsUKvLy8uPjii9u1b0LYnKZpLXkJISws+/qIFvt/67TSiqp2r+vvXyVq/f66Tvv9nXdp/fv31yorK6vn3XPPPVpUVJRmNBobXf7DDz/UnJyctJMnT1ZPW7VqlaaU0pKTkzVN07SioiKtpKSk1nIXLlzQvL29tcWLF9eabt7WK6+8oumnCCG6tEZzl9TgRJc1Z84cRo8ezbp16xg0aBBeXl5cddVV5OTkcPz4caZMmYK3tzejR4/mwIED1csZjUaWLVtGv379cHd3Jy4ujnfffbfWutetW8cVV1xBaGgofn5+jBs3jm+++aZWmcWLFxMcHMzevXsZN24cXl5eJCQk8NNPP7E7JZfBEX54uDq3ez+HRQZQadBY99XXzJo1CxeXmkHjs2fP5uzZsxw6dKjR5b/++mvGjBlDnz59qqf9+te/xs3NjfXr1wPg7e2Np2ftzjBBQUHExMSQnp5ea7o0Swp7IQlOdGmpqaksXLiQJ598kn//+99s3bqVefPmMXv2bGbPns3q1aupqqpi9uzZaJoGwP3338+TTz7JvHnzWLduHTNnzmTu3LmsXbu2er2nTp3immuu4f333+fTTz/l4osvZsaMGWzZsqXW9ktKSrjjjju4++67+fTTT3F3d2fWrFnsO5VZ3TxpNBqpqqpq8mUwGBrdx6GR/hgryshMP0t8fHyteQMHDgQgKSmp0eWTkpLqLefm5kZsbGyTy2VnZ3P8+HHi4uIaLSNEd9b6+wsJ0YlycnLYtm0bsbGxABw4cIDnnnuOd999l9tvvx3Qm9mvuuoqkpKScHV15fXXX+ftt9/mjjvuAODyyy8nIyODJUuWcPXVVwNw3333VW/DaDQyZcoUDh8+zH/+8x8uueSS6nmlpaUsX76cqVOnAhAeHk5CQgJOJw8y8paLAJg7d269GmJdkydPZuPGjQ3O6xXkiY8qByAgIKDWvMBAPYnm5jZ+t5Pc3Nx6y5mXbWq5hx9+GB8fH+bMmdNk7EJ0V5LgRJfWu3fv6uQG0K9fP4DqhGM5LS0tjRMnTuDk5MTMmTOpqqqqLnPZZZexcuVKDAYDzs7OnD17lgULFvDdd9+RkZFRXfuzTG6g14QuvfTS6t8HDRoEQFXh+eoa3OLFi2slzIb4+vpWvzcYDNXbA3BxcWFguC9Hmj8cVvP666/zwQcf8Omnn9KjR8ueRC5EdyMJTnRpdWsmbm5u9aabp5WVlXH+/HkMBgP+/g3f3T8jI4OIiAiuvfZaCgsLeeKJJ+jXrx/e3t4sXLiQc+fO1Srv6+uLk1NNS755W74uRiL8PQCIjo4mKiqqyf2wvK4VGxtLSkpK9e+nTp0ioV8knwHnL9SucZlrYOaaXEMCAwPJz68/ji43N5fhw4fXm/7ll19y//3388wzzzBz5swm4xaiO5MEJ+xKUFAQLi4ubNmypVZiMgsNDeX48ePs3buXr7/+munTp1fPKy0tbfF2egV5VSet1jZRrlmzhvLy8up5ERERjCp0x9k3hK17DnDn72uWM19Dq3uNzVJ8fHy9a20VFRWcPHmS+fPn15q+ZcsWZs+ezfz583nkkUea3U8hujNJcMKuTJ06FYPBQH5+PldccUWDZcyJzN3dvXpaSkoKW7ZsYdiwYU2uP7tQT0xRgV7V01rbRDl06NB684dG+ePZdxTr136J4eXncXbWe2euWrWKXr16MWTIkEbXPWPGDD788ENSUlKIiYkB9FpaeXl5rQR++PBhrrnmGqZPn97iAeRCdGeS4IRdGTBgAPPnz2f27Nk8+uijjB49mrKyMg4fPkxycjIrVqwgPj6eqKgoHn74YZYuXUphYSGLFi0iMjKy2fXvSdWbDHsF1XS57927N717925X3JEBnkRfejPH/vUHbrvtNu666y527tzJG2+8weuvv16ridPFxYWFCxeycOFCAG644QaeeuopZs2axdKlS8nPz+ehhx7i1ltvpX///gCcO3eO6dOn4+Pjwx//+Ed27NhRvT4/P7/qa4ugDzsoLi5m3759AKxevRqAMWPGVCdQIboDSXDC7rz22mvExcXx5ptvsnDhwuoT+O9/r7f9ubu789lnn3Hvvfdyww03EBUVxYIFC9i4cWOT482gJsGF+7f9BssNUUoxetggfOY9y/Ht7zFjxgzCwsJ44YUXuPPOO2uVNRgMGI01Tx9wdXVl/fr13Hfffdx00024u7sze/ZsnnvuueoyiYmJnD17FoApU6bUWl/dHp733HNPrWuEN954IwBvv/229LgU3Yqy7M3VhBYVEsLe3fSvbVQYjHxx7yXNF26lF745yj83nuDQ4ml4urV/ALkQDqLROxPIQG8hWqjSYORAWtuf4N2cIZH+GIwaiRkFHbJ+IRyNJDghWuhIRgFllUZGxgR0yPqHRelDG+TROUJYhyQ4IVpoT4p+/S2hg2pwYX4eBPu4c+CsJDghrEESnBAttCc1j55+7tUDvK1NKcWwKH8OpuV1yPqFcDSS4IRooT2puYyMDuzQu+0PjfTn+LkiSiqqmi8shGiSJDghWuBcYRlnc0s7rIOJ2bAof4waJKZLRxMh2ksSnBAtsCclD6DDOpiYDY3UO5rIdTgh2k8SnBAtsDc1F1dnxeCIhm/ibC2hfh709HPnwNm8Dt2OEI5AEpwQLbAnNZfBEf5WeYJ3c0ZGB7I7tfHnuAkhWkYSnBDNqKgycuBsfodffzMb3TuIMzmlZOaXdcr2hLBXkuCEaMaRjALKqzpugHddo00PUt2VktMp2xPCXkmCE6IZ5hssd1YNblCEH56uzuw6Lc2UQrSHJDghmrEnNY8wPw8iAqzwBIF9K/VXE1ydnUiIDpAanBDtJAlOiGbsScllVIyVam/5Z/RXM8b2CeJwegG5xRXW2a4QDkgSnBBNyCooIy2vlITogE7d7qS4EDQNfj5+vlO3K4Q9kQQnRBPMN1i2Wg2uhYZHBeDv6crm5OxO3a4Q9kQSnBBN2J2Si5uLU4cP8K7L2UkxoX8wm49l08KHEgsh6pAEJ0QTdqfmMizSHzeXzv9XmRwXQlZBOUezCjt920LYA0lwQjSirNLA4bSCTm+eNJvUPwSAjUelmVKItpAEJ0QjDqfnU2EwdtgDTpsT5u/BwHA/vj+SZZPtC9HdSYITohGd9QSBplw5qCe7U3K5UFRusxiE6K4kwQnRiN0puUQHeRHq2zFP8G6JKwb1xKjB90nnbBaDEN2VJDghGqBpGrtTcxnZyePf6hoc4UdkgCffJkozpRCtJQlOiAaczS0lu7DcZh1MzJRSXD4wlJ+OZVNaYbBpLEJ0N5LghGiA+QbLtupgYunKwWGUVRr56Zj0phSiNSTBCdGAPSm5eLo6Ex/ma+tQGNsnCD8PF2mmFKKVJMEJ0YADafkMjfTHxdn2/yKuzk5MiQ/l+6RzGIxyVxMhWsr2/71CdDGVBiOJ6QUMi+rc23M15YpBPckprmB3ijwjToiWkgQnRB3JWYWUVxkZ2oUS3OS4EFydFd8mZto6FCG6DUlwQtRx8Gw+oN/Rv6vw9XDl4thgvk3MkpsvC9FCkuCEqGP/2Xx8PVyI6eFl61BquWJQT05fKOH4uSJbhyJEtyAJTog6DqblMSzKH6WUrUOp5fKBPQG5+bIQLSUJTggLZZUGkjIKGdaFmifNwvw9iA7yYldKjq1DEaJbcLF1AKIb2/tf+N8fGp434Cq45UP9/Y9Pw6Zlja8nZgL8bl3btgPgHw0PHdTfp/4C6x+D7KMQEA2XL4IBM2rK5p+FV0bDsBvh2lfqrSops5Aqo8awyHZ2MCnJgR+ehKR1UJoDATEw8jbQjKDa/r1ydExg9UNQG61hvjUdUrfp7/tMhju+rJm3byUkrYXMA1CUDe4+EDoQJj0CfSY1H8DJjbDlZTh3BEougFL6vsVfBRMf1tcHUFUO3y2GQ5+CoRLipsGMZ8DD4riu+zPseQ/u/QWC+rTlcAjRJKnBCdtz87bO8qV5sHI2FGfDbz8FF3f4+HbIOVlT9rsl4OQCUx9vcFUHz+YBMKxXQNvjqSiBd66GXf+BokwwVMCFY/DtQjj6ddvXC4zqHcj5ogpSLpQ0XGD/RzXJrSE/vaAnuLxUqCrVj9WpzfDuNXBwdfMBpO2BE99DYToYyqGqDM4fhZ9fhE/uqCm35SXY/k8YeQdc9jjsXwkbFtTMz06G3W/DuPmS3ESHkRqcaD/LGlRDpvyf/rL0y7/h60f090NvbHr9Cb/RX5aOroeVN5uWv0H/eXanXlsaPhtixsOwm2HD/8GJHyGoL6TthoOf6LU6n9AGN7X/bD49vN2I8G/FEwTS9kBlCfSeYNq31+HcYf39ZYv0k/yaP+qJJWMfhI9o+bqzkyH3lF4DAkbHBAH6kw56B9f5YlBWoCdRVy89noZ4+MPUv+nHxjMQfv6HnvQANj9XcywbEzYMbnwXel0EngH6sf1kjp7sjn+n11y9giB5vV7+4vv0bX63GJI31Kznm7/p25/455YfCyFaSWpwwjZ2v63/9OoBg65r/fK73tJ/OrnCyNv194YK/aezq/7Txa329A0LIDAGxjXe3HnwbD5DW9LBpCQHfnkDXr8E3pwCp36yWImpJuTmC5c8AN49YMKfauZnHWp63RXFsPcD+M80eG0MHP6iela/UB88XJ04nF5Qf7mNy6AoCyY1kTRu/5/eHBkQDe6+ek3W3U+fZ1nTbUz/y2Hwr8EvHFw9If5XEBpfM9987Kv/Fm41Pw2mZ9qd3AjHNuiJ1sOv+W0K0UZSgxPtV5gBz/SG8iL9xDnkev0k6+LecPnU7XAuUX8/4jc1iail8s7A8W/19/FX1dTGIkaCiwcc/0G/HpS8AVAQPV5PEqnb4Kb3Go2rpKKKY+cKmTYkrOHtahqc2gR73ocja0wnbAW9J0LfS/UylWWQnaS/D+wNTs76+x6xNespauSekmd3w5534dBnUFFYs08Dr6ku4uykiA/z43B6fu1lzx2BHW9Aj34w/n74/omGt2G+RmZmqACj6SkFvuENL9OYylI48QOcM+3v0Jv0pAkQcwlkHtSPU0C03hQ64FdgNOpfNHoOhYTbW7c9IVpJEpxoP2MllJpuIZVzAjY/qzcH3vZZw+XNtS8UjP5d67e35129swbAmN/XTPcL1zuOrP0TLIvWaw1XPAGhg/TrQzETamqLhsqa2obJkYwCjBoMrdvBJD8N9n0Ie9+HvBR9Wo/+MPxmGDYbAnrVlC3Lq4nN3eJGze4WNZWK4pr3JTn6dbO979ckfb8oGHsnDL8VQuLq7f7gCD++3J9eu6PJV4+AsQpmPNu6LwxbX4ZKUzwjb2vZMuVF8HRk7WkDr4HrXqv5ffJf9KT72V367+HDYfoyfT+zDsEda8DJ1IDUwN9CCGuQBCfarkcsXPsq9J0M3iFwdhesngvF5/SOCKd+gj4Tay9TkgOJ/9Pfx07Rr421hqFKr0GBnmTq9vwbdhMMngUFZ/UaiYu73usvL1WvvaXvgzUP6L0I3Xxh1B1w+RJwciLR1Ow3OKJOs9nb0/XlPQJg9O9hxK0QNbp1cWNx9xHL5s+Pb4fTP4GrNwy/Rb9+2HtSzcm/AUMi/fnvL6mcySkluoeX3iR6+ic9yfS7rOUh7VsJP/5df997IlzyYOt2ydKRNXpP1+tX6L97Bem9N4sv6F+AfMP0xPjjUxB/tf5l49uFsOsdvbYaNgyueQkiRrQ9BiHqkGtwou2ix+nf+gOi9esxfSbqveLM0vfUX2bfh3rPO4DRc1u/zaNf6T0TofHan7OL3jzo4q6fYH96Xk9KIQNh1W/hfDLMfENPzFtfhj3vAJCYUUCAlyvhdTuYKFMzo6ECygugLF9vamuIZ2DNMIDywprplu9dLe6QYi5rqNA7iZQV6DWxJpgT8CFzM+Xm5/X1DL8VMg7oL7OKYv338jp3P9n3oZ6QNKPehHvLypbXotx9YHE+/DUD5qzTa5ygd+BJ31u7rHcPPbmB3qGlNBeuXKof8y0v6X+Dmf/W/yarboOqipbFIEQLSIITbdfgSV418t5k9zv6T98IiJtRf35zzM2bLp560mrOxqf1OKcu1Lvq55+BvlP0mp65M8aJHwA4nF7AoHC/+h1M5v0IM57Ta6wHP4EPZsE/BsE3j0NWYu2yLu4QYup0kXu65vrWhRM1ZXx61ry/ZSVc90+9Rnh0HXx8G7wQpzeznt3V4C7F9fTF2UnVXIerKNYT1Ue3wBsT9ZdZ2i79d8vEs/e/8L979WX6TNKHVFg2p7aUm5fec3TQtTXTLPfTUv5Z2PYaXHS3Xms3HXMmPaKPSex7KeSn6n8jIaxEEpxou5WzYfu/9E4flWV6k+T212vmR4+rXf7U5poT2Mjb9ZpWXT8+DYv99VduSu15OSf1HngAg2fqtaWmmMdaTXwIfHvW1MTM23Uy1VicXKgyGEnKLKzfPAn6di6aB/N/hnmbYMydejf8rS/D6+PhXxNr4oKarvYVhXotpfiCPk7MrOeQmvdu3voQiLnr4b7deq9LZzd9DN2Ky+CVUfXGp3m4OhMb4s3RzEJabe8H8OV9enLrdznc+knD4xBzU2r+Dj8+XTN93Z/15FSUrf/NU7dDosVA8sBGxrR9t0TfzqRH9d+r/xY1f4NaP4WwAvk0ibYrSIf1f9FfdQ29EXqNrT3NXPtSzvq1r9ba/Q7V17Ja0rz5zd/ALwLG36f/3qOf/jq5SU+2hz7Vp8dN50R2MRVVRgY1lOAsRYzQX1c+BUe+1O/EcfpnSNlW05Ny3B/g4Kf6WLjvl+gvs/AR4B9Zf70Awf30TjFTF+rd6Pe8r/cWPf59vfFp/UN9a2pwDY1BXGzqKFP3TiYbn6npBHP8O3iqZ+3lHjigD6VozIGPYeebDc8b8CuIGlV/unn84dUv1gwLGDADEr+AHf/Wv6yc2qT/bYJi6y8vRBtJghNtN+WvcGi1PtC5MFPvPBEcBwm/1TtjWCrKhiNr9fdx0/XE0xpVFXrTGkDYUOg1puny5rFWN75TMyzA2QVmf6j3OFx5i14zm/o4DJ9N4t6zAAyOaOEtulw99GbOYTdBzil9gHn1PE+Ys9Z0q661+nWngGi91lpZ2vy6nV304Q/xV+nH9cLxekViQ334+lAGZZUGPFydWxazNYy9Ux/cnXtavxbp5gMhA/ShIWN+3/AyGxboPVlHWnypGXYzFKTpnUwOfAyRI+FXzzdcqxeijVQLny0lD6ASdu3JtYm8vz2Fw0um4eLcgS33m57Vf05+tF2rWbM/nftX7uXrByYyMFwGSwuH1uhdGeQanBDoPSjjw3w7NrlZUb9QfcD2MXk2nBCN6h7/zUJ0IE3T9B6UzV1/60L6BHvjpJCHnwrRBElwwuGl55eRX1rJoJZef+sCPFydiQ7y4vi5NvSkFMJBSIITDu9wmt4bcVA3u5bVL9RXanBCNEESnHB4iRkFKAUDw9sw2NmG+oX6cOp8MVWGRu6qIoSDkwQnHN7h9AL6BHvj5da9uqj3D/Wh0qCRktPIs9+EcHCS4ITDSzTdoqu7iTX1pDwhzZRCNEgSnHBo+SWVpOWVtnyAdxfSN0S/xdaJ7OJmSgrhmCTBCYd2OMPUwaQbDREw8/NwJdTXnRPZUoMToiGS4IRDS8rQu9l3tw4mZrEhPpyUBCdEgyTBCYeWlFlAD283QnzcbR1Km/QN8eZEdjEtvOWeEA5FEpxwaEczCxkQ5lv/GXDdRGyID/mllVwolgeFClGXJDjhsIxGjeSsIgaEdc/mSZCelEI0RRKccFipOSWUVhqI784JztST8uR56UkpRF2S4ITDSjI9ETs+rPv1oDSL8PfEw9VJanBCNEASnHBYSZn6LbrienbfGpyTk6JPsI8MFRCiAZLghMM6mllITJAXnm6d+ETsDhAb4i1NlEI0QBKccFjmHpTdXWyID2dySiirNNg6FCG6FElwwiGVVRo4faG4W19/M+sb4o1Rg5QLctNlISxJghMO6VhWEUaNbt2D0iw2xDRUQK7DCVGLJDjhkI5kFgDYRROl+abLcssuIWqTBCcc0tHMQjxcnYjp4W3rUNrNy82FyABPeaqAEHVIghMO6WhmIXE9fXF26p636KpLvyel1OCEsCQJTjikpMxCBnTj8W91xYb4cOJckdx0WQgLkuCEw8ktruB8UXm3HuBdV2yIN8UVBs4Vlts6FCG6DElwwuEkZ+m36Orf08fGkVhPdU9KuWWXENUkwQmHk2xKAvZUg+srQwWEqEcSnHA4x7IK8XV3Idzfw9ahWE1PP3e83ZylJ6UQFiTBCYeTnFVIv54+3fYhpw1RShEbKjddFsKSJDjhcI6fKyIu1H6aJ80G9PTlSEahrcMQosuQBCccSk5xBeeLKuyqg4nZoAg/zheVc66wzNahCNElSIITDqWmB6X91eAGhus3jk5ML7BxJEJ0DZLghEM5ZkpwcXZYgzMnOGmmFEInCU44lOSsInzdXQjzs58elGb+nq5EBXqSmCE1OCFAEpxwMMlZhfS3sx6UlgaG+5GYnm/rMIToEiTBCYdy7FyRXQ3wrmtQuB+nzhdTWiFP9xZCEpxwGOeLyskprrDLDiZmgyL8MGpwNEuuwwkhCU44jGQ77mBiNjhC72hyME2aKYWQBCccxrEs+7sHZV2RAZ4E+7ixLzXP1qEIYXOS4ITDSM4qxM/DhVBfd1uH0mGUUozoFcDeM7m2DkUIm5MEJxyGuYOJvfagNEuIDuRkdjH5JZW2DkUIm5IEJxyCpmkcMw0RsHcjegUAsP9snk3jEMLWJMEJh3C+qILckkr62+FNlusaFuWPUrBXrsMJBycJTjiEmlt02X+C8/VwpX+oD/vkOpxwcJLghENwhCEClkb0CmDfmTw0TbN1KELYjCQ44RCSzxXh7+lKiB33oLQ0olcguSWVpFwosXUoQtiMJDjhEI5lFRJnx/egrCshOgBAhgsIhyYJTtg9TdNIziqy61t01RXX0xcvN2cZ8C0cmiQ4YfeyC8vJL60kLtQxrr8BODsphkcFsPdMnq1DEcJmJMEJu5fsALfoasjImAAS0wsoq5QnCwjHJAlO2D3znfUdqYkSIKFXIFVGTW68LByWJDhh95IyCgj2cXeYHpRmI8wdTVKlo4lwTJLghN07klnAwHDHqr0BBPu4Ex3kJXc0EQ5LEpywa1UGI8lZRcSHOV6CA324gCQ44agkwQm7dup8MRVVRgaG+9k6FJtI6BVAZkEZGfmltg5FiE4nCU7YtSOZegeT+DAHTXDRgQDsScmzbSBC2IAkOGHXkjIKcHFSxIZ62zoUmxgY7oe7i5N0NBEOSRKcsGtHMgroF+qDu4szX3zxBcOGDcPd3Z0+ffrw4osvNrnsQw89hFKKP//5z7WmJyUlcdFFF+Hv78/s2bMpKiqqNX/z5s1ERkbWm96Qd955B6VUg2UXL15McHBw9e+nT59GKVX98vb2JjY2lt/85jf89NNP9ZafM2cOF48by9BIfxnwLRySJDhh15IyC4kP82XLli3MmjWLsWPHsmbNGubOnctf/vIXli9f3uByiYmJ/Oc//8HPr37T5pw5c+jXrx8ff/wxiYmJ/P3vf6+eZzQaeeCBB3j66afx8emYO6c8//zzbNu2ja+++orHH3+cCxcuMGnSJJYsWdJg+YToAA6m5VNRZeyQeIToqiTBCbuVV1JBRn4Z8eF+PPHEE1xyySWsWLGCK6+8kscff5w//vGPPPHEE1RUVNRb9v777+eBBx4gMDCw1vSioiJ++eUXli9fzrRp01iwYAHffvtt9fy33noLV1dXbrvttg7brwEDBjBu3DgmT57MnDlzWL9+PY8//jiLFy9m48aN9conRAdSUWXkSEZBh8UkRFckCU7YrUTTCT0+zJd9+/ZxxRVX1Jp/5ZVXkpuby7Zt22pNX716NUlJSTz22GP11mlOhp6engB4eXlVTysoKOBvf/sbL730Uqc/tWDRokVERETwr3/9q94885MF9sh1OOFgJMEJu3U4TU9wgyP8KSsrw83NrdZ88+9HjhypnlZaWsrDDz/MsmXL8Pau3zElKCiI3r1788orr5CTk8O///1vRo8eDcDSpUu5/PLLGT9+fKtjNRgMVFVV1XoZjS1vUnR2dmbq1Kls37693rxwf09CfN05lCY1OOFYXGwdgBAd5WBaPuH+HoT4utOvXz927txZa/6OHTsAyMnJqZ729NNPEx4ezm9/+9tG1/vPf/6TG2+8kb/+9a/079+f1157jePHj7NixQoOHjzYplgDAgIanN6jR48WryMqKoqsrKwG5w2J8ONwutyTUjgWqcEJu3UwLZ+hkf4AzJ8/ny+++II333yT3NxcNmzYUN2L0slJ/zc4deoUzz//fLNNjDNmzODcuXMcPXqUI0eOEB0dzZ/+9CceeughoqKieO2114iOjiY6Opp//vOfLYp18+bN7Ny5s9brrrvuatX+aprW6Lwhkf4cO1ckTxYQDkVqcMIuFZRVcup8MdePjARg7ty57N+/n3vuuYd58+bh5eXFM888w/33309YWBgAjz32GDNmzGDAgAHk5eUBeq/I8vJy8vLy8Pf3r058Xl5exMXFAfDtt9+yf/9+Vq1axf79+3n88cfZunUrAOPHj2fChAkMGzasyXgTEhLq9bpcu3Ztq/Y5LS2Nnj17NjhvcIQ/BqNGUmYhI3oFtGq9QnRXUoMTdumQ6RExQ0w1OGdnZ1599VWys7M5cOAAWVlZjBs3DqD659GjR/nss88IDAysfp05c4ZXX32VwMBA0tLS6m3HYDDw0EMP8eyzz+Lp6cnGjRuZOnUq8fHxxMfHc9lll7Fp06YO39+qqip++OGHRq//DYnUhzsckkfnCAciNThhl8wncnMTpZk5cYF+Le3iiy8mPj4egBUrVtQbcD179mwmT57MPffcQ0hISL3tvP766wQGBnLzzTdXTyspKal+X1xc3GTTobU88cQTpKenM3/+/AbnRwZ4EuDlKtfhhEORBCfs0oGz+UQGeNLDR38G3Pbt2/n5558ZMWIEBQUFrFy5kg0bNvDzzz9XL2PuDWnJw8ODXr16cemll9abl5OTw5IlS9iwYUP1tEmTJvHoo4/y1ltvoWkaP/zwA8uWLbPqvh09epTg4GAqKio4deoUH330EevXr2fx4sVMnjy5wWWUUgyJ8JeelMKhSIITdumQRQcTAFdXV1atWsXixYtxcnJi4sSJbNmyhaFDh7Z5G4sXL+baa69l5MiR1dMSEhJ49tlnWbBgAaDfdWT48OFt35EGmG8d5uHhQXh4OOPHj2fz5s1MnDixyeUGR/rx9s+nqagy4uYiVyeE/VMtbD7p+DYWIawkv6SS4U98wyPTBnDvlH62Dqe2Tc/qPyc/2umbXrM/nftX7mXdHycwOMK/+QWE6B4a7fIsX+OE3Tlkus40LEpO4pbMHW4OSzOlcBCS4ITdOWjuQSm1lFpigrzwcXepPj5C2DtJcMLu7E3NJaaHF4Hebs0XdiBOTopBEX7VNVwh7J0kOGFXNE1jd0oeo6IDmy/sgIZE+HMko4Aqgzw6R9g/SXDCrpzNLeV8UTkjYyTBNWRIpB9llUZOZBfbOhQhOpwkOGFXdqfoj4QZJQmuQeaON3IdTjgCSXDCruxOycXH3YW4nr62DqVL6hPsg5ebs9yySzgESXDCruxOyWVErwCcnTr3gaPdhbOTYnCEn9TghEOQBCfsRnF5FUmZBXL9rRlDIv05nJ4vHU2E3ZMEJ+zG/jN5GDW5/tacoZH+0tFEOASr3ouy0lDJS3teYl/2Po5cOEKFsQKAd6e/y8ieI5tZWoj22Xk6F6WQ5501w7KjyYAwuVYpbCe1IJW3Dr3FvnP7OJl/Eg0NZ+XMvtv3WWX9Vk1wpYZS3k1815qrFKLFtp+8wKBwP/w9XW0dSpdm2dHkhlFRtg5HOLBjecf49NinHbZ+qzZRuigXZg+YzbKJy7gp7iZrrlqIJpVVGtiTmsu4vj1sHUqXJx1NRFfR06sndw29i1envsrQ4LY/2aMxVq3Bebl6sWCc/piQlIIUa65aiCbtP5NHeZWR8ZLgWmRoZAAf7kih0mDE1VkuxQvbGBI8hCHBQwB45/A7Vl+/fLKFXdh28gJKwZg+QbYOpVsYFRNIWaWRw+nyZAFhvyTBCbuw/eQFBkfI9beWGt1b72m663SOjSMRouNIghPdnn79LY9xfaR5sqV6+nkQHeTFTklwwo616RrczsydzN0wt9a0pZcs5df9fm2NmIRolb2peVRUGRkfKwmuNUb3DmTT0Ww0TUMpufOLsD9SgxPd3pbj53F2UnL9rZXG9A7iQnEFp87LgG9hn9pUgxsTNoaDdxxscF5umX4397KqsuppRZVF5Jbl4u7sjperV1s2KUSjNiVnMzI6AD8Puf7WGmOqr8Pl0jfEx8bRCEdUaaykqKIIgCpjVfV0cx7xdvXGzbntDy626jABgEmrJtWbdu/39wJwbey1PDXhKWtvUjiw80XlHEzL589Xxtk6lG4nNsSHIG83tp44z01jetk6HOGA9p3bV+9yl0EzVOeR9l76kiZK0a39dCwbgMlxoTaOpPtRSjGpfzCbj53HYNRsHY4QVmf1GlxjTZdCdIRNR7Pp4e3G4Ag/W4fSLU2JD+WLfekcOJtHQrTcpFp0rqYud1mD1OBEt2U0amw+dp5JcSE4yfPf2mRS/xCcFPx4NNvWoQhhdZLgRLd1MC2fnOIKJseF2DqUbivQ242E6EB+SMqydShCWJ0kONFtfZuYhZOCSZLg2mX64DAOpRXIcAFhdyTBiW5rw+FMLurTgyDvtncjFnDN8AiUgi/2ptk6FCGsShKc6JZOZBdx7FwR04eE2TqUbi/M34NxfXrw5f50NE16Uwr7IQlOdEsbDmcCcOXgnjaOxD7MTIjk1Plidp7OtXUoQliNJDjRLW04lMnwXgGE+3vaOhS7cPXwcPw8XHh322lbhyKE1UiCE91Oel4p+8/mM32wNE9ai5ebCzeP6cX6Q5lk5JfaOhwhrEISnOh21uxPB+BXQyXBWdNt43pj1DT+uz3V1qEIYRWS4ES38/neNBKiA4jp4W3rUOxKdA8vLh/Ykw9+SaGkoqr5BYTo4iTBiW7lSEYBSZmFzEyItHUodmn+5L7klVTy8c4ztg5FiHaTBCe6lS/2peHipLhqaLitQ7FLo2KCGB0TyJs/naLKYLR1OEK0iyQ40W1UGYx8sTeNSXEh9PBxt3U4duvuybGk5ZWy7mCGrUMRol0kwYlu4/ukc2QVlDNbnl3WoS6LD6VfqA//2nRSBn6Lbk0SnOg2PvwllTA/D6bGy7PfOpKTk2LepL4cyShg87Hztg5HiDaTBCe6hTM5JWw+ls1NY3rh4iwf24523YgIevq588amE7YORYg2kzOF6BY+2pmKAm6W5slO4e7izO8n9GHriQscOJtn63CEaBNJcKJFNE3j73//O7169cLT05NJkyaxb9++JpcxGAw888wzTJw4kR49etCjRw+uvPJKdu7c2eRyM2fORCnFq6++CkBZpYGVO84wNT6UyAC5NVdnuWVsNL7uLryx6WT1tP/9738MHToUDw8PBg0axKpVq5pdzyeffMK1115LZGQkPj4+jBo1ipUrV9Yrt2rVKmbNmkV4eDhKKd555x1r7o5wQJLgRIssW7aMpUuX8pe//IU1a9bg4+PD5ZdfTmZmZqPLlJaWsmzZMsaMGcP777/PBx98gKurKxMmTGD37t0NLvPNN9+wbdu2WtM+3XOWnOIK7pzY16r7JJrm6+HKb8bF8PWhDE6fL+bnn3/m+uuvZ8qUKXz99ddcddVV3HLLLXzzzTdNrufFF1/Ex8eHf/zjH3z55ZdMmTKFW2+9lVdeeaVWudWrV3P69Gmuvvrqjtwt4Ug0TWvJSziw0tJSzc/PT1uyZEn1tKKiIi04OFhbsGBBo8tVVVVpOTk5taaVl5drMTEx2pw5c+qVr6io0OLj47UVK1ZogPbKK69oBoNRu/S5H7VrXvlJMxqN1tspW9n4jP7qJrLyS7X+f/1KW/D5Ae3KK6/UpkyZUmv+jBkztEsuuaTJdWRnZ9ebdsstt2i9e/euNc1gMGiapmmFhYUaoL399tvtC144ikZzl9TgRLO2bt1KQUEBN910U/U0b29vrrnmGr7++utGl3N2diYwMLDWNDc3NwYPHkx6enq98i+99BKenp787ne/q572TWImp84Xc9fEviilrLA3ojVC/TyYNTKSVdtP8eOPP9b6DADMnj2bbdu2kZ+f3+g6goOD601LSEio9xlwcpLTkbAu+USJZiUlJeHs7Ez//v1rTR84cCBJSUmtWld5eTl79uwhLi6u1vTMzEyWLl3K8uXLq090mqax/Ltj9A32ZoY82NRm7prUl5ILaVRWVhIfH19r3sCBAzEajSQnJ7dqndu2bav3GRDC2iTBiWbl5ubi4+ODs7NzremBgYGUlJRQUVHR4nU99dRT5OTkcN9999Wa/uijjzJt2jQmTZpUPe1wun7fyQcu7y9DA2woNsSH0WFuALh7+dSaZ66h5+a2/EGp33//PV988QUPP/yw9YIUogEutg5AdB2apmEwGKp/V0rVS2rtsW7dOp566ileeOEFBgwYUD1927ZtrF69miNHjtQqv/HoOfpN8OHqYRFWi0G0zXUjIvgSWH84i/Fj276e06dPc+utt3LdddcxZ84ca4UnRIPka7GotmnTJlxdXatfl112GaB/Sy8qKqqV/ED/1u7l5YWbm1uz6965cyc333wz8+fP58EHH6w178EHH+Tuu+/G39+fvLw88vLyAMjKLeSui3ri7CTX3mxt3MAYAFZvPUpFVc1NmM01t7rXWhuSk5PDjBkziImJ4b///W/HBCqEBanBiWqjRo2qNUbN19cXgPj4eAwGA8ePH69V80pKSqp3TaYhycnJXHXVVVx22WW8/PLL9eYfPXqUHTt2sHz58lrT8za+zW8mv8fNVfJsMluLjY3FxdWVzDMnWbUzldvG9wb0z4CTk1Oz19NKSkq4+uqrqaioYO3atXh5eXVC1MLRSYIT1Xx9fRk9enS96RdffDF+fn588skn/O1vfwP0E9aaNWuYN29ek+vMyMhg2rRpxMbGsnLlygabPNeuXUuVRRJ7f9tp3vrr77h17t3Mu/2Wdu6VsAZ3d3emTpnCvtO/8PIPx7l+VBRebi6sWrWK8ePH4+/v3+iyVVVV3HjjjRw7doytW7cSGir3EhWdQxKcaJaHhwePPfYYS5cuJTAwkPj4eF588UWMRiP3339/dbn33nuPuXPncuLECWJiYigtLWXGjBnk5uby6quvcuDAgeqy7u7uJCQkADBhwoTq6fvP5LHxmxIAxicMYfLkyZ20l6I5jz/+OJdeeimln7/CX72zcEvfz1dffcX69eury6SkpBAbG8tbb73F7bffDsAf/vAHvvrqK1566SUuXLjAhQsXqssnJCTg7q4/+igxMZHExETKysoA2LVrFz4+PoSEhMjnQLRNU4PkNBnoLUyMRqP25JNPapGRkZqHh4c2YcIEbc+ePbXKvP322xqgnTp1StM0TTt16pQGNPiKiYmpt43Siirtshc2auP+/l31QG+7080Getf1+eefawGRfTXl7Kr1j4vTVq5cWWu++W9uOUg7Jiam0c+B+bOiaZq2aNGiBstMnjy5c3ZOdFeN5i6ltex5T/JQKNHhFv3vEO9uS+G9uWOZFBdi63A6xqZn9Z+TH7VtHO2QmF7AVa/8xO3jYlhy3RBbhyNEo73QpBel6BI+33uWd7elcOeEPvab3OzEoAg/bh8Xw/vbU+RJA6JLkwQnbC4xvYD/++wgF/UJ4rEZzffKFLb38LQB9PBx56+fH6w1bECIrkQSnLCprIIy7npvF34errxya4LcsaSb8PNwZel1gzmUVsCL37buNl1CdBY5mwibKSyr5I63dpBXUsFbc8YQ6uth65BEK0wfEs4tY3vxxuYTbDjc+GOThLAVSXDCJsoqDdz9/m6Onyvi9d+OYkhk4+OoRNe18OrBDI8K4I8r97I7peX3oxSiM0iCE52urNLAXe/tYtvJCzx7wzDpVNKNebo58587RhPm78Gct3ew41SOrUMSopokONGpzMnt5+Pnefb6YcwaGWXrkEQ79fBx58O7xhHi685v//ML7287TQuHHwnRoWwyDm757uXWXJ3oJkrKnVm7JZz0Cx5cMSaLwb0LbR1SpxubvBGAHXGX2jSOjlBa7sSGHWGczvSmb0QRl406h7eHofkFhcN5cNSD1lxdo+PgbHKrruN5x22xWWFDBYW+7No/krJyN0YO24d7QCbH82wdVefrV6Y34dnr/8DgIcl4+fUm6Xgc73wdxaC4JKIi0pCHsQtbkHtRig5lNCpOpPQh+UR/3FwrGD/6FwL9820dluggSkHfmNOEBmdzIHEI+xOHkZYZwdCBh/D2KrV1eMLBSIITHaKy0oX0rHBOnO5DSak34T0zGBp/GDe3SqttY8e7O9j5/s7mC3YhgZP0Z+e9tvknG0fSuDG3jWHsHe14qing413M+NG/kHq2F0eOxbNp20Ti+yXTJ/q01OZEp5EEJ6yistKF3PwA/ZUXSE5eIEajM36+BYwZsYueIdm2DlF0MqUgptcZQkOyOXhkMInJA0nLiCC+/1GCgy5IohMdThKcaBVNg5JSLwoKfSko8qWwyJeCQj9KSs0PsNTw8ykkJuoMkWFp+PsVyInMwXl6lDFmxG7Ss8I5kjyAX/aMxc83n8iwDIICc/D1KcLFWTqjCOuzSS/K+76/z5qrE1ZyISeIikpXjJoTmlFh1JwwGp0oK3enpNSLkhIvCot9MBjM34s0vL2K8fMtxM+3gED/PAL88nFxkZNVY6an6s/EWx89zMaR2IbB6MSZtCjOpEeSXxBQPd3ZuQpX10qclIYyvQCGxh+mR5CMrbM3r172qjVX17V6UfYL6GeLzYpm7NjVi3O59W+XpZSGr1cV/t6V9A4tIti/nOCAcnr4VeDqYv7u4wQEmV6iMa6FRYBj/w8MCAKGZlNYkktmjgd5Ra6UljtTWu6MpoFRU5i/d/cJDCc0QD5Tom3keXCi2vFzRVQajLg6O+Hm7ISLs8LV2YlAL1e5CbIQoqtqtAYnCU4IIUR3Jg88FUII4VgkwQkhhLBLkuCEEELYJUlwQggh7JIkOCGEEHZJEpwQQgi7JAlOCCGEXZIEJ4QQwi5JghNCCGGXWnQnkyVLlqwHgq243Qgg3Yrr62wSv21J/LbV3eOH7r8PEn+N84sWLZre4BxN0zr9tXjxYs0W25X4Jf6u8JL4bf/q7vsg8bfsJU2UQggh7JKtEtwSG23XWiR+25L4bau7xw/dfx8k/hZo6dMEhBBCiG5FmiiFEELYJUlwQggh7JIkOCGEEHbJ6glOKTVPKfWjUipPKaUppXq3cLnrlVKJSqly08+ZdeYrpdRipVS6UqpUKbVRKTW4A+J3V0q9opQ6r5QqVkp9qZSKamaZ06Z9rftaZ1FmcQPzM7tI/M3G1sWP//8ppXYqpQqUUtlKqTVKqSF1yrzTwD5ut1LMf1BKnVJKlSmldiulJjZTfrKpXJlS6qRSan5719lZ8SulZimlvjEd50Kl1C9KqWvrlJnTyP+DRxeI/9JGYouvU67J85EN42/oc6wppYpbu49Win2S6X80zbSNOS1YZqhSapPpPJKmlFqolFJ1yljn+Ft73AHwIPB/pp8a0LsFy4wHqoAFwEDTzyrgIosyfwEKgeuBIcDH6AMFfa0c/+um9V4BjAQ2AvsA5yaWCQHCLF4JgBG4w6LMYiCpTrmQDjj+bYm/2di6+PHfAPzOFNdQ4HMgEwiyKPMO8G2dfQyyQrw3A5XAXabP7itAERDdSPk+QLGp3EDTcpXA9W1dZyfH/xLwGDAW6AcsAgzARIsyc0z7aHmsw6wdexvjvxT9vDSoTnzOFmWaPR/ZMH7/uscVOAG83Zp9tGL8vwL+DtwAlABzminvZ/rf/Nj0/3oD+nnl4Y44/lb/wFkEOZqWJ7hVwLd1pn0HrDS9V0AGsMBivqfpwNxtxZj9gQrgNxbTeqEnq2mtWM8CIA/wtJi2GDjUUce7PfE3F1s3PP4+6CfdayymvQOs7YBj/gvwZp1px4CnGyn/DHCszrQVwLa2rrMz429kHTuAFyx+nwMUWTtWKx3/S03npeAm1tnk+agrHX/gEtP+XNyafeygv0URzSe4e4CCOufGvwFp1PTqt9rx7yrX4MYD39SZtgG42PS+D/o3kOoymqaVApstyljDKMC1znbOAEdauh1TVfv3wAemGC31NTXxnVJKfaSU6muluM3aE39TsXWb42/ii978nltn+gSl1DmlVLJS6k2lVGh7glVKuZlirvvZ/YbG423ssz5aKeXaxnW2iRW35Uv9Y+2plEpRSp1VSq1VSiW0I9QGtTP+XUqpDKXU90qpKXXmNXc+sgorHf+7gMOapm1tYF5T+2gr44Gf6pwbN6Dfuqu3RRmrHP+ukuDCgKw607JM07H42VQZa8VhAM63YztXoCeEN+tM/wX9m+109A9lGLBVKdWjrcE2oK3xNxdbdzr+oDej7QO2WUxbD9wOXAY8jN7E9oNSyr2twaLfn9WZ1h2Xxj7rLqb1tWWdbdXubSml7gWigPctJh8F5gLXAbcAZcAWpVT/9gZcR1viz0CvRVwPzDLF+n2d617NnY+spV3HXynlD9xE/XNNS/bRVho7tuZ5TZVp9fF3aUkhpdST6M1uTZmiadrG1gbQGVoav5U2dxewU9O0/ZYTNU37uk5M24GTwB3Ai02tsKPjb09sLdGZx18p9SIwAZigaZrBPF3TtI8sih1USu0GUoCrgM+ssW1Ho5S6HngOuFnTtBTzdE3TtmHx5UIptRX9C8f9wB87OcxaNE07in7CN9um9I5wjwA/2SSotvsteiXF8suFve1ju7QowQHLgQ+aKZPajjgygZ51pvU0TcfiZ88627Es05TltCz+cejfqIKB7DrbafaDYWryug64t7mymqYVKaUOAy35VrucToi/idi6y/H/BzAb/cvWyabKapqWrpQ6S8uOf2POo9c4m/rs1tXYZ73KtD7VhnW2VVviB0ApdQPwHnC7pmlrmiqraZpBKbWL9h3rhrQ5/jp+Qf/cmDV3PrKW9sZ/F/Cppmk5LShbdx9tpbFja57XVJlWH/8WNVFqmnZe07SkZl4lrd24hW3oTXuWrgDM7cqn0Heuuoypy/FEizLWiH83eo8my+1EoffkaXY76M185cDK5gqa4o9Hb07oKvE3FluXP/5KqZfQm8OmapqW1IJ9DAYiacHxb2K/KkwxN/XZrauxz/ouTdMq27jONmnrtpRSN6HXGuZomra6ue2YrksPox3HuiFWPFYjqB1bc+cjq2hP/EqpscBw6jdPNmYEVj7+bbQNmFhnyMgV6D2nT1uUsc7x74CeNGHoB/NW9J48vzL9btll+3ssegmhXzysQu9+HI8+zKCS+sME8tHblIcAH9Fx3dTPApejd/f/kTrd1NG71N9XZzkFJFOnR5TF/OeByejX5y4C1qL3Joqxdfwtia0rH3/gNVO8U6ndLdrHNN/HtI/j0S9kX4r+T3S2vfGjd/OuAO5ET8QvofcmizHNfw94z6K8eZjAclP5O03L1x0m0Og6rXy8Wxv/bPT/zQfqHGvL/+9FwDSgL/r//lumZcZ2gfgfBH6NXpscDDyNfp6aZVGm2fORreK3WG4FkNzIOpvdRyvG72P6G49AHyaw0PQ+2jT/aeB7i/L+6F+WP0I/j8xC/9+1HCZgteNv1Z01BbfYdDDrvuZYlDkNvFNnuRvQT1wV6L3mZtWZr0zrzkC/aL0JGNIB8bujj0W5YPqDrQF61SmjAYvrTJtimt7gPzE1CaECvUvsp8CgrhB/S2Lryse/kc9bdRn0IQ0bgHOmfUxBHzbQy0ox/8H0mS5H/0Y+yWLeRmBjnfKTgT2m8qeA+a1ZZwcc8xbHb/q9oWNtWeYfpmNcbjrmG4DxXST+R9G74ZcCOehN379qYJ1Nno9sFb9pmi96Eny0kfW1aB+tFPuljXwe3jHNfwc4XWeZoeg9sMvQzyeLMA0RsPbxl6cJCCGEsEtdZZiAEEIIYVWS4IQQQtglSXBCCCHskiQ4IYQQdkkSnBBCCLskCU4IIYRdkgQnhBDCLkmCE0IIYZckwQkhhLBL/w/msP3GKGg8QAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "idata = az.from_pymc3(trace, model=ordinal_model)\n",
    "az.plot_posterior(idata, var_names=['b'], ref_val=0, rope=[-1, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 788,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:Auto-assigning NUTS sampler...\n",
      "INFO:pymc3:Initializing NUTS using jitter+adapt_diag...\n",
      "INFO:pymc3:Multiprocess sampling (4 chains in 4 jobs)\n",
      "INFO:pymc3:NUTS: [b, a_offset, stratumid_sigma, intercept]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='32000' class='' max='32000' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      100.00% [32000/32000 06:03<00:00 Sampling 4 chains, 0 divergences]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:Sampling 4 chains for 4_000 tune and 4_000 draw iterations (16_000 + 16_000 draws total) took 365 seconds.\n",
      "INFO:pymc3:The number of effective samples is smaller than 25% for some parameters.\n"
     ]
    }
   ],
   "source": [
    "import theano.tensor as tt\n",
    "import pymc3 as pm\n",
    "\n",
    "stratumid = d.stratumid.astype('category').cat.codes\n",
    "treatment = np.repeat(d.treatment.values, 4).reshape(-1, 4)\n",
    "P = 80\n",
    "y = d.timeseekhelpfever.cat.codes\n",
    "levels = 4\n",
    "\n",
    "with pm.Model() as categorical_model:\n",
    "    intercept = pm.Normal('intercept', mu=0, sigma=1, shape=levels)\n",
    "    stratumid_sigma = pm.HalfCauchy(\"stratumid_sigma\", 2.0)\n",
    "    \n",
    "    a_offset = pm.Normal(\"a_offset\", mu=0, sigma=1, shape=(P, levels)) \n",
    "    a = pm.Deterministic(\"a\", a_offset*stratumid_sigma)\n",
    "    \n",
    "    b = pm.Normal(\"b\", mu=0, sigma=1, shape=levels)\n",
    "\n",
    "    eta = intercept + a[stratumid] + b*treatment    \n",
    "    p = pm.Deterministic('p', tt.nnet.softmax(eta))\n",
    "    yl = pm.Categorical('obs', p=p, observed=y)\n",
    "\n",
    "    trace = pm.sample(draws=4_000, tune=4_000, return_inferencedata=False, target_accept=0.9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1129,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(16000, 4)"
      ]
     },
     "execution_count": 1129,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trace['b'].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 789,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([<AxesSubplot:title={'center':'b\\n0'}>,\n",
       "       <AxesSubplot:title={'center':'b\\n1'}>,\n",
       "       <AxesSubplot:title={'center':'b\\n2'}>,\n",
       "       <AxesSubplot:title={'center':'b\\n3'}>], dtype=object)"
      ]
     },
     "execution_count": 789,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1987.2x331.2 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "idata = az.from_pymc3(trace, model=categorical_model)\n",
    "az.plot_posterior(idata, var_names=['b'], ref_val=0, rope=[-2, 2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 790,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5900625"
      ]
     },
     "execution_count": 790,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "((trace['b'][:, 0] > 0) & (trace['b'][:, 3] < 0)).mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 728,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "summary = az.summary(idata)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 733,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='854' class='' max='4000' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      21.35% [854/4000 00:35<02:12]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "with moo:\n",
    "    post_checks = pm.sample_posterior_predictive(trace)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 734,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "idata = az.from_pymc3(trace, model=moo, posterior_predictive=post_checks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "post_checks['obs']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "y_pred = post_checks['obs'].mean(0)\n",
    "\n",
    "jitter = np.random.normal(0, 0.03, len(y))\n",
    "\n",
    "plt.figure(figsize=(12, 5))\n",
    "plt.scatter(y + jitter, y_pred, alpha=0.4)\n",
    "plt.xticks(range(4), d.timeseekhelpfever.unique())\n",
    "plt.yticks(range(4), d.timeseekhelpfever.unique())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 867,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 63.,  31.,  31.,  13.,  12.,  63.,  63.,  63.,  63.,  63., 127.,\n",
       "       127.,  63., 127., 127., 127., 255., 255., 255., 255.])"
      ]
     },
     "execution_count": 867,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trace.tree_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 831,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "with m:\n",
    "    prior_checks = pm.sample_prior_predictive(samples=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 827,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0         More than 2 days\n",
       "1       Less than 24 hours\n",
       "2       Less than 24 hours\n",
       "3        Did not seek help\n",
       "4       Less than 24 hours\n",
       "               ...        \n",
       "1278    Less than 24 hours\n",
       "1279      More than 2 days\n",
       "1280    Less than 24 hours\n",
       "1281    Less than 24 hours\n",
       "1282      More than 2 days\n",
       "Name: timeseekhelpfever, Length: 1283, dtype: category\n",
       "Categories (4, object): ['24-48 hours' < 'Less than 24 hours' < 'More than 2 days' < 'Did not seek help']"
      ]
     },
     "execution_count": 827,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "d.timeseekhelpfever"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['userid', 'stratumid', 'wave', 'language', 'week', 'month',\n",
       "       'survey_duration', 'survey_start_time', 'survey_end_time', 'shortcode',\n",
       "       'admalaria', 'admnm', 'age', 'buymosquitonet', 'caste', 'castespecify',\n",
       "       'consent', 'default_tys', 'distancemedicalcenter', 'dwelling',\n",
       "       'education', 'familymembers', 'fever4months', 'gender',\n",
       "       'hasairconditioning', 'hasmosquitonet', 'introendline',\n",
       "       'knowbuymosquitonet', 'knowmnm', 'longsleeves', 'malaria4months',\n",
       "       'membersbednet', 'membersnotbednet', 'mobilecurrent', 'mobilenumber',\n",
       "       'mobileoperator', 'notifyme', 'numbermalaria4months',\n",
       "       'numbermosquitonet', 'occupation', 'paymentprocess', 'paymosquitonet',\n",
       "       'pregnantwoman', 'religion', 'religionspecify', 'seekhelpfever',\n",
       "       'sleepundernet', 'sprays', 'talkmalaria', 'testmalaria',\n",
       "       'timeseekhelpfever', 'timeseekhelpmalaria', 'treatmentseeking',\n",
       "       'treatmosquitonet', 'typemedicalcenter', 'wait', 'worriedcovid',\n",
       "       'worriedmalaria'],\n",
       "      dtype='object', name='question_ref')"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 772,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from bambi import Model\n",
    "\n",
    "model = Model(ddb[ddb.seekhelpfever.notna() & ddb.dwelling.notna()].reset_index())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 758,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:bambi:Modeling the probability that seekhelpfever==1\n",
      "INFO:pymc3:Auto-assigning NUTS sampler...\n",
      "INFO:pymc3:Initializing NUTS using jitter+adapt_diag...\n",
      "INFO:pymc3:Multiprocess sampling (4 chains in 4 jobs)\n",
      "INFO:pymc3:NUTS: [1|stratumid_offset, 1|stratumid_sigma, treatment, Intercept]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='16000' class='' max='16000' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      100.00% [16000/16000 00:26<00:00 Sampling 4 chains, 4 divergences]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:pymc3:Sampling 4 chains for 2_000 tune and 2_000 draw iterations (8_000 + 8_000 draws total) took 28 seconds.\n",
      "WARNING:pymc3:The acceptance probability does not match the target. It is 0.886277293636303, but should be close to 0.8. Try to increase the number of tuning steps.\n",
      "ERROR:pymc3:There was 1 divergence after tuning. Increase `target_accept` or reparameterize.\n",
      "ERROR:pymc3:There were 3 divergences after tuning. Increase `target_accept` or reparameterize.\n",
      "INFO:pymc3:The number of effective samples is smaller than 25% for some parameters.\n"
     ]
    }
   ],
   "source": [
    "results = model.fit('seekhelpfever ~ treatment + (1 | stratumid)', family='bernoulli', draws=2_000, tune=2_000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 768,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:title={'center':'treatment'}>"
      ]
     },
     "execution_count": 768,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "az.plot_posterior(results, var_names=['treatment'], ref_val=0, rope=[-1, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "results = model.fit('seekhelpfever ~ treatment*pucca + (1 | stratumid)', family='bernoulli', draws=2_000, tune=2_000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 774,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([<AxesSubplot:title={'center':'treatment'}>,\n",
       "       <AxesSubplot:title={'center':'treatment:pucca'}>], dtype=object)"
      ]
     },
     "execution_count": 774,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA+kAAAFTCAYAAABF801FAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Z1A+gAAAACXBIWXMAAAsTAAALEwEAmpwYAACK1UlEQVR4nOzdd3hUZdrH8e+Z9J5AQihJaKH33juC2LvYe13Xta+7NmxrWfVV17KuDbsoKgpKkyYgvbfQS0ggIb23mXn/OEkIJSRAkjPJ/D7XNddkzpxyn5kkz3Of8xTD6XQiIiIiIiIiItazWR2AiIiIiIiIiJiUpIuIiIiIiIi4CCXpIiIiIiIiIi5CSbqIiIiIiIiIi1CSLiIiIiIiIuIilKSLiIiIiIiIuAgl6SI1yDCMSwzDeMiiY7cyDGOSYRhtrDh+ddWXOEVExDWobK1afYlTRKpHSbpIzboEsKQiAbQCngFcvYBuRf2IU0REXMMlqGytSivqR5wiUg1K0kUsYBiGj9UxiIiINCQqW0WkoVCSLlJDDMOYDNwEtDAMw1n62GcYxsjSny8zDONDwzCOAEkVtrvTMIwNhmEUGIaRYhjGx4ZhNDpu3/cZhrHMMIw0wzAyDMNYbhjG+RXeHwksKH05t8LxR5a+v88wjC8Nw7jBMIzthmHkG4ax2DCMdoZhBBiG8YFhGKmGYSQZhvG6YRiexx0/wjCM/xqGkWAYRqFhGHGGYdx53Do3lx5zoGEYXxmGkWUYRqJhGG8bhuFbnThFREQqUtlaO2VradN4p2EY3QzDWGAYRp5hGIcMw3jOMAxbhfXKjt/qZNsft8zTMIy/G4axtfRzP2IYxizDMDoed87vGYYRX3rO8YZhfGGUXmAxDCO29PXe0s9zj2EY7xuGEVbZuYg0RJ5VryIi1fQ8EAH0Ay4qXVYIhJT+/B9gJnADUFawvgw8DLwNPAq0AF4AuhqGMdjpdNpLt20FfATsw/y7vRCYYRjGBKfTOQtYC/wFeBe4H1hVut3WCvENB9oCfwe8gTeBH4A9wC5gYuk6TwK7gfdKYwwGlgB+wCRgLzAeeN8wDB+n0/mf4z6HL4BvgMuAQaXbpGM2w6tOnCIiImVUtprOuGw1DGNS6XqtnU7nvuP2Ow34BHip9PhPAY7S/Z+ubzG7JrwJ/I75fQwHmgFxpYn2n0AjzO9jI9AEuBjzsysEmgPxwAOl59cG+CfwW+l5i7gHp9Ophx561NADmAwcPG7ZSMAJ/HTc8laAHXj6uOVDSte/pJJj2DArE3OAn09ynLEn2WYfkAaEVFh2f+n6Hx237lpgQYXXTwEFQLvj1vsQSAE8S1/fXLq/Z49bbwawozpx6qGHHnroocfxD5WtZ1e2Ak8DJUDLCssmla7/+EmOnw2EHnf8VsetNwlwVng9unS9+0/xPT5X+t30Oo3v3hMYWrrvam+nhx71/aHm7iJ156fjXp+DWSn4qrSJmGdpU7gVmAXk8LIVDcPoYxjGDMMwkjAL2uLS7TucxvGXOZ3OzAqv40qfZx+3XhwQXeH1uaUx7T0uztlAY6Dzcdv/etzrTUDMacQpIiJSXSpbq+B0Op9zOp2eTqdz/0ne/u64198CgUDX6uy7gnGYifSHVayzyul0rqtsBcMwvA3D+Gdp0/98zO9kcenbp/O9iNRrau4uUncOHfe6SenzrkrWbwxgGEY0MA+z2dpfgQOYlYnngU6ncfz0414XnWK573FxxmIWlJXGWUHaca8LAQ3mIyIitUFl69lJquR1i9PcT2Mgzel05lexzoYq9vMS5vfxHGbT+GwgCviRYz8/kQZNSbpI3XEe9zq19HkcJxbmFd8/F7Pv3VVOp/Ng2ZuGYfjXeIQnlwokA3+r5P3tdRSHiIjI8VS2np1IzP7zFV8DJJQ+F5Q+ex+33fEXEVKARoZh+J0iUU+h6uR/IvC50+l8oWyBYRiBVWwj0uAoSRepWYWYg8BUx1zMwVlinE7n3FOsV1ZhKL/abhhGe8z+dQcrrFdY+lzd41fXLErvMjidzuQa2F9txSkiIg2TytaqnWmcVwEvV3g9EcjBbE4PUNZEviuwA8xR3DEvglQ0B3gcuB1zML+TmQM8aRhGD6fTWdkddX9ObF1wSxXnINLgKEkXqVlbMa8k3wOs5ugV6BM4nc7dhmG8ArxjGEYHYFHp+tGYfeI+cjqdCzBHSC0BPjcM43XMUVKfxWyaV3FciR2l691qGEYaZoG93el0Zp/lOf0fcDWw2DCM/8O8uh8AdASGOZ3Oi09zf7UVp4iINEwqW6tWaZyGYTyNOXhc25P0S7+jdMq1VZiju98OTKrQz34V5qj0/y5drxC4l+Oa2judzgWGYfwAvFHalWA+4IU5BsCvTqdzYek5Xwv8bhjGC5gXAsIxR3e/u/QznQXcZBjGJswuC5cBg0/zsxCp9zRwnEjN+ghz0JV/ASuB6ada2el0/hO4E7MQ+w74GXMal3RgZ+k6W4DrgJbAL8BjmFer/zhuX6nAfUAPzErJKqDP2Z5QaUE9GHP6k79jDmrzCWahuuAUm1a2v1qJU0REGiyVrVXv71Rx2gAPwDjJphdjXrz4Bbgec2q05yvst6R0nXjMUfbfxWytMPkk+5qIOer7JaX7+wToQum4AU6nMwOzpcJPmJ/1LOB1zIsLZX35/1q67YvAFCAIuKZaH4JIA2I4ncd35RERERERkYaqwtzpXqWJuIi4EN1JFxEREREREXERStJFREREREREXISau4uIiIiIiIi4CN1JFxEREREREXERStJFREREREREXERV86SrLbyIiIjrOdlUSiejclxERMT1nLIc1510ERERERERERehJF1ERERERETERShJFxEREREREXERStJFREREREREXISSdBEREREREREXoSRdRERERERExEUoSRcRERERERFxEUrSRURERERERFyEknQRERERERERF6EkXURERERERMRFKEkXqScKiu0kZRVgdzitDkVERERcULHdQXJWARl5RaoviNRjnlYHICKnlpiRz6uz4vht82GKShwE+Xhyae8W/G1MOxoH+lgdnoiIiFikoNjO3K1JzNuWxKp96SRk5Je/5+Vh0DsmjLGdIrm6fzTBvl4WRioip8NwOk95lU2X4EQstPZAOrdNXkVBsYMr+0bRrkkg6w5k8MuGRCKCfPj4pn50bh5sdZgiUveMaq6nclykAdpzJIcPF+9hxoZDZBeWEB7ozYDWjWkfGUSjQG+KSxwkZuTz5+5Uth7KIsjHk3tHxXLn8DZ42Kr770NEatEp/xCVpIu4qL0puVz0nyU0CvTm05v70SYisPy9zQmZ3P7ZarILipl8a3/6tWpkYaQiYgEl6SJuaH9qLm/9vpNp6xPw9rRxfrfmXNa7BQPbNK40+d50MJO35u3k921J9G/diP9c04vIYN86jlxEjqMkXaS+KSyxc/n7f3IwPZ9f7x9Gi1C/E9ZJyirgmv8t50hOIVPuHKQ76iLuRUm6iBvJLijmnQW7+GTJXjxsBjcMbMmdw9sSEVS9bm9Op5Mf1ybw1M+baRTgzRe3DaB1eEAtRy0ip6AkXaS+eWnmNj5YtIcPb+zLOZ0jK10vISOfK97/kxKHk5/uHUxUmH8dRikiFlKSLuIGHA4nP6w9yKuzt3Mku5Ar+kTx2PgONDnDO+EbD2Zw86ersBkw9e7BtFKiLmIVJeki9cmu5GzOfXMxl/eO4pUrule5/o6kbC5//0+aBvsy9Z7BhPhpYBgRN6AkXaSB25GUzWNTN7I+PoOe0aFMuqgLPaNDz3q/u5JzuPK/fxLq782P9wwmLMD77IMVkdN1ynJcU7CJuBCn08mz07fi5+3BY+d2qNY27SOD+OD6PuxLzeXuL9ZQVOKo5ShFRESkttgdTv67aDcXvL2EA2l5vH5lD368Z3CNJOgAsU0C+d+NfUlIz+eer9ZoqjYRF6QkXcSFLN6ZwuKdKTw4tv1pTa82ODacly/rzrI9qTw2dYMKXBERkXpoz5Ecrvjvn7w8M47RHZsw58HhXN4nClsNj8jer1UjXry0K8v3pPHegl01um8ROXuaJ13ERTidTt6at5PmIb5cNzDmtLe/vE8Uh7MK+Pfs7Tic8PpVPfDy0HU4ERERV+d0OvlqxQFe+HUrPp4evDWxJxf1aI5h1N50aVf0iWLxzhTenLeTwbHh9GkZVmvHEpHToxq8iIv4c3cqa/anc8/Itvh4epzRPv4yKpa/n9uRXzYkcu2Hy0nIyK/hKEVERKQmZRcU89dv1vHktM30a9WIOQ8O5+KeLWo1QQcwDIMXLu1K02Bf/v7DRnWXE3EhGjhOxEVc87/l7EnJ4Y/HRp1xkl5m2roEnvhpEwDXDojhkl4t6BAZhIfNIC23iH2peRxMz+NIdiGpuUX4enrQIsyP4e3Cz3jEWBGpUxo4TqQB2JKYyV++Wkt8ej4Pj2vP3cPb1njT9qosiEvmlsmreHR8B/4yKrZOjy3ixjS6u4ir25qYxXlvL+YfEzpy14i2NbLP/am5vD5nBzM2JuJwgs0w/6CP/5P3tBmUVOjDfk7nSB4b34F2kUE1EoeI1Aol6SL13MxNh3jwu/WE+nnz9jW96N+6kWWx3PPlGubHJfP7QyOIbqTpXEXqgJJ0EVf36PcbmLHxEMv/MYYQ/5qdQu1QZj4r96axKzkHgFB/b1qH+xMd5k+TYF+CfT0pcTjZlZzDzE2H+HTpPvKL7Tx7cReuG9CyRmMRkRqjJF2knnI6nfx30R5emRVH75hQ/ndjX8JPY7DY2nA4s4BRry1kbOdI/nNNL0tjEXETStJFXFlKTiGDX5rP1f2ief6SrlaHQ2pOIQ99t4FFO45w1/A2PD6hY633ixOR06YkXaQeKrY7ePKnzUxZHc+FPZrz7yu64+t1dl3casrrc7bzn/m7+PkvQ+hRQ9O9iUilNE+6iCv7avkBiuwObh7SyupQAGgc6MMnN/fj+oExfPDHHv67aI/VIYmIiNR7BcV27vlyDVNWx3P/mHa8PbGnyyToAHcOb0PjAG9emrmNKm7iiUgtU5IuYqGiEgdfLN/PqA4RtI0ItDqcch42g+cu6sqFPZrzyqw45m5NsjokERGReiu/yM4dn6/m923JvHBJVx46p73LtVIL8vXib2PbsXxPGgu3H7E6HBG3piRdxEJzth4mJaeQmwa3sjqUE9hsBq9d2Z0uzYN5bOoGkrIKrA5JRESk3skpLOGmT1eydFcKr13Zg+sHuu54L9f0j6F1eAAvz4zD4dDddBGrKEkXsdA3Kw/QItSP4e0irA7lpHw8PXj7ml4UFDt4bOpGNX8TERE5DZn5xVz/0QrW7E/nrYm9uKJPlNUhnZKXh40HxrZje1I2c9SKTsQyStJFLLIvJZelu1K5pn90nc+JejraRgTy6PgOLNpxRM3eRUREqimnsIQbP17B1sQs3r+uNxf2aG51SNVyQffmtGrsz7sLdunivIhFlKSLWOTbVfF42Ayu7BttdShVumFQS9pHBvL8r1spKLZbHY6IiIhLKyi2c8dnq9mcmMV71/VmXJemVodUbR42g3tGtmVTQiZ/7EyxOhwRt6QkXcQCRSUOpq6JZ0zHJkQG+1odTpW8PGxMurAL8Wn5fLFsv9XhiIiIuKwSu4P7v1nHsj2pvH5lD8Z2jrQ6pNN2aa8omof48u78XVaHIuKWlKSLWOD3bUmk5BRxzYAYq0OptsGx4QyNDee/i3aTV1RidTgiIiIux+Fw8viPm5izNYlJF3bmkl4trA7pjHh72rhzeBtW7ktj5d40q8MRcTtK0kUsMHXNQZoG+7rsgHGVefCcdqTmFuluuoiIyHGcTicv/raNqWsO8uDY9tw8pLXVIZ2Vif1jaBzgzQeLdlsdiojbUZIuUseSswpYtOMIl/VugYcLDxh3Mn1aNmJYu3D+98ce9U0XERGp4OMle/l4yV5uHtyK+8fEWh3OWfP18uCGQS2ZF5fM7iM5Vocj4laUpIvUsWnrE7A7nFzu4tOwVOaeEW1JzS3ilw2JVociIiLiEmZtPsyLv21jQtemPH1BZwyjfl2Er8z1A1vi7WnjkyV7rQ5FxK0oSRepQ06nk+9XH6R3TChtIwKtDueMDGrbmI5Ng/hkyV5NzSIiIm5vfXwGD0xZR4+oUP7v6p4uPa3q6QoP9OGyXi2YuuYgablFVocj4jaUpIvUoY0HM9mZnMMVfVx/2rXKGIbBrUNaE3c4m2W7U60OR0RExDLxaXnc/tkqIoJ8+Oimvvh6eVgdUo27dWhrCkscfLVc49GI1BUl6SJ1aOqag/h42ji/ezOrQzkrF/VsTqi/F1+tPGB1KCIiIpbIzCvmlsmrKCpx8OnN/QgP9LE6pFrRPjKIEe0j+GzZfgpLNB6NSF1Qki5SRwqK7fyyIZHxXZoS4udldThnxdfLg0t6tmDuliTS1fxNRETcTFGJg7u/XMP+1Fw+uKEvsU2CrA6pVt02tDUpOYX8tumQ1aGIuAUl6SJ1ZN62ZDLzi7ming4Yd7yr+0VTZHfw07oEq0MRERGpU89O38KyPam8cnl3BrVtbHU4tW5obDitwwP4XFOwitQJJekidWTqmniahfgyJDbc6lBqRKdmwXSPCuG71fEaQE5ERNzG1ysO8NWKA9w1og2X9W4YF96rYrMZ3DCwJesOZLDpYKbV4Yg0eErSRepAfZkbvbCwkIcffpgmTZoQEBDA+eefz759+ypd/8o+UcQdzmZ7Ujb79+/nmmuuoVGjRvj7+9OjRw9mzZp10u1yc3OJjo7GMAw2b95cS2cjIiJSs1bvS+OZXzYzon0Ej43vaHU41bZ06VIGDBiAr68vrVu35u233z6t7XNzc3l64nD2v3IBr0/5/Zj3pkyZwmWXXUazZs0wDIPJkyfXYOQi7klJukgd+GVDIg4nLn/F/f7772fy5Mm89tprTJ06lZSUFM455xwKCgpOuv6Ebs2wGfDF72sYNGgQGRkZfPrpp/zyyy/ccMMN5Ofnn3S7F198keLi4to8FRERkRp1OLOAu79cS4tQP96e2MulL7pXtGvXLsaPH0/r1q357bffuOuuu3jooYf46KOPqr2PF198kZISs9xeuD35mPFopk6dyr59+7jgggtqPHYRd2VU0UxVbVhFasCF/1mCYcAv9w21OpRKHTx4kFatWvHJJ59w4403ApCQkEDr1q157733uP3220+63fUfrWDOO/+gQ1AxixYtwmY79bW/Xbt20bNnT1577TXuueceNm3aRNeuXWv8fEQauOpmByrHRWpAQbGdqz9Yxq7kHKb9ZQjtIuvPQHF33XUXCxYsYOvWrXh6egJw7733Mn36dA4cOIBhnPrfyfHldrNb3+GZG8/lrhFtAXA4HNhsNnJycggKCuLTTz/l5ptvru3TEqnvTvmHpzvpIqfp5ptvpm/fvvz666907twZf39/zj//fNLS0ti1axejRo0iICCAvn37snHjRnYfyWFTQiYXdGvKyy+/TGxsLD4+PrRv357PPvvsmH3/+uuvnHPOOTRp0oTg4GAGDhzInDlzjlln0qRJhIeHs27dOgYOHIi/vz+9evVi8eLFZ3VeZce57LLLype1aNGCoUOHMnPmzEq3G9M2gCObFnPB1TdXmaADPPDAA9x+++107Fh/mgmKiEjDcbrlOMAzP29hfXw6A7MXM2FIL5csxyszc+ZMLrvssvIEHWDixIkcPHiwWl3Oji+3u7UI4csV+7E7zGuA1Sn7ReT06K9K5AwcOHCAp59+mhdeeIH//e9//Pnnn9x5551MnDiRiRMnMnXqVEpKSpg4cSI/r0vAMGDZl6/xwgsvcOedd/Lrr79y6aWXcuuttzJjxozy/e7du5cLL7yQL774gh9++IHBgwczYcIEli5deszx8/LyuOmmm7jrrrv44Ycf8PHx4bLLLiMvL698HYfDQUlJySkfdvvR+U7j4uKIiooiMDDwmGN16tSJuLi4Sj+L8IJEcJSwLj6DIUOG4OXlRVRUFC+99NIJA8r9+uuvLF++nGeeeeaMPncREZGacDrl+I9r45myOp4WW79hyodvuWw5fjK5ubnEx8efcGG8U6dOAKcs3+Hk5fZFPZoTn5bPoh3Jp/6QReTMOZ3OUz1E5Dg33XST08PDw7lr167yZY8++qgTcH722Wfly3799Vcn4Oz3yGTn+c9/5zQMwzl58uRj9nXDDTc4+/bte9Lj2O12Z3FxsXPcuHHOW265pXz5M8884wSc8+bNK1+2bt06J+CcOXPmMXFiNnWt9DFixIjy9W+//XZnjx49TojjiSeecDZr1qzSz+Prr792Ak4PnwDnY4895pw/f77zqaeectpsNue7775bvl5hYaEzNja2fNmCBQucgHPTpk2V7ltEKlVV+a1yXKQSp1uOt7n7A+eE51y/HD+ZgwcPOgHnTz/9dMzy4uJiJ+D84IMPKt22snJ77foNzn4vzHXe+PGKY9bPzs52As5PP/30lDGJiNPprKL8PtruRUSqrVWrVrRt27b8dWxsLACjR48+Ydn++IN097Bhs9m49NJLKSkpKV9nzJgxfPPNN9jtdjw8PDh48CBPPPEEv//+O4cOHSq/Ez1kyJBjju/t7c3IkSPLX3fu3Bkw+5WXmTRpEvfdd98pzyMo6Oz71JXF6NO6N1fd+zh9WoYxatQoDh48yEsvvcS9994LwBtvvIGvry933XXXWR9TRETkbFSnHI+KaQ2AkZ/OyKAi5rh4OW63249pwVaxefuZqKzc9vKwce2AGN78fSf7UnJpFR5wVscRkRMpSRc5A6Ghoce89vb2PmF52TIPRzHNfG3Y7XZCQkJOur9Dhw7RvHlzLrroIrKzs3nuueeIjY0lICCAp59+muTkY5uUBQUFHdMHrOxYFUdhj4mJISrq1KPJVxwsJiwsjMzME+c+TU9PJywsrNJ9lL0X0KoHc7Yepk9L8/Xo0aP59NNPycrKorCwkBdffJHJkyeTnZ0NQE5ODgDZ2dnk5uYSEKBCXkRE6kZ1yvH3F+8D4LaBLShO3e/y5Xjbtm3Zv39/+eu9e/cSEREBcEL5np6eDlBp+X7kyJFTltsXd2nPO/N38cXy/Tx1QedTxigip09JukgtKRtQpVtUCM0jffH09GTp0qUnHWClSZMm7Nq1i3Xr1jFz5kzOPffc8vcqm8asKrfeeusJA9ocb8SIESxcuBCAjh07Eh8ff0LCHBcXd8pB3sr6tbVq7M+8bcn8Y4L5uuxqvs1mIyEhgZycHK644ooTth88eDBjxozh999/P+E9ERERK8zecpjpGw8B0C0qlHhntsuX49OnT6ewsLD8vebNm+Pt7U10dPQJfc/LXldWvlen3D731lf5fnU8j4zrgJ+3x+mcmohUQUm6SC1Zf8C8Sj2gdSNGjx6I3W4nMzOTc84556TrlxXiPj4+5cv279/P0qVL6d69+2kf/3SbyY0bNw6An376ieuvvx6AxMREFi9ezHvvvVfpPlq1akWXLl0gcTO7ooazNyWX1uEBzJs3j7Zt2xIYGEhsbCwLFiw4Zrv169fz4IMP8sknn9C7d+/TPj8REZHakJZbxBM/bSI2IpCE0mWjR492+XK8W7duJ11nwoQJ/PTTT7zwwgt4eJjJ9JQpU4iOjq50CtTqlNuFwdHM2HiIXzYkcHW/mNM5NRGpgpJ0kVoyd2sSAD2iQunQoQN33303EydO5LHHHqNv374UFBSwZcsWduzYwUcffUTHjh2Jiori4Ycf5vnnnyc7O5tnnnmGFi1anNHxW7VqRatWraq9flRUFLfddhsPPPAATqeTiIgIJk2aRMuWLcuTdoDnnnuO55577pg+ec8//zyXX345Qbm+/KdJKn6p2/niiy/4/PPPAQgMDDym711F/fr10zzpIiLiMp75ZQuZ+cW8fHEnFj1vLqsP5XhlHn30Ub766ituuOEG7rjjDlatWsUHH3zA+++/f0xzeU9PT55++mmefvrpapXbTqeT9pGBfDBtER77/cqb6q9evZrAwEAiIiIYMWLEWccv4o6UpIvUgsISOwtKpybx8TKvWr/77ru0b9+eDz/8kKeffprg4GA6d+7MbbfdZq7n48OPP/7IX/7yF6644gqioqJ44oknWLhwYbXmMa0Jb7/9NgEBATz00EPk5eUxYsQIvvnmG3x9fcvXcTgcJ0z5cumll/L5559z98NP8NZjv9C6VUveffddrrvuujqJW0REpCbM3nKY6RsSeWRce2KbeB3zXn0ox08mNjaWWbNm8dBDDzFhwgSaNm3K66+/zu23337Mena7HYfDUe39GobBtf1jeODvH3Dlv74pX/7uu+/y7rvvHtMUX0ROj1FxFMiTOOWbInJys7cc5q4v1jD5ln6M7NDE6nDqzOtztvPewt2seXIsof7eVocj0pAZVa8CqBwXqbaUnELG/d8ftAj146d7B+PpcWLfczlWZl4x/f/1O5f3ieJfl568ub2InNQpy3H99xGpBb+sT6RxgDdDYsOtDqVOjekUid3hZOH2I1aHIiIiclqen7GVnIISXr+qhxL0agrx9+KC7s35eV0COYUlVW8gItWi/0AiNSy3sITftyVxXrdmeLlZId+9RQgRQT7M3ZZkdSgiIiLVtnxPKj+vT+TuEW1oHxlU9QZS7toBMeQW2fllfaLVoYg0GO6VQYjUgXlxyRSWOLigezOrQ6lzNpvByPYRLNmZQom9+v3aRERErFJsd/DMz1toEerHPSNjrQ6n3ukdE0rHpkF8vXJ/1SuLSLUoSRepYb9tPEREkA99WzWyOhRLjOgQQWZ+MRsOZlodioiISJU+X7af7UnZPH1hZ833fQYMw+DaATFsTshi48EMq8MRaRCUpIvUoNzCEhZsT2ZC16Z42Ko7rlPDMjQ2HJsBi3aoX7qIiLi2lJxC/m/uDkZ2iGBc50irw6m3LunVAj8vD75eccDqUEQaBCXpIjVofmlT9/O6uV9T9zKh/t70jA5Vki4iIi7vP/N2kl9s56kLOh8zZ7icnmBfLy7s0YxfNiSSXVBsdTgi9Z6SdJEa9NumQ4QH+tDPTZu6lxnRvgkbD2aQlltkdSgiIiIndSA1j69XHuDqftG0jQi0Opx679oBLckrsjNNA8iJnDUl6SI1JK9ITd3LjOgQgdMJi3fqbrqIiLim1+dux8Nm8Lcx7awOpUHoERVC52bBfL3iAE6n0+pwROo1JekiNWRB3BEKit27qXuZbi1CCPP3UpN3ERFxSVsSM/l5fSK3DGlNZLCv1eE0CGUDyG07lMX6+AyrwxGp15Ski9SQsqbu/Vu7d1N3AA+bwbB2EfyxI0VX00VExOW8PmcHIX5e3D2irdWhNCgX92yOv7cGkBM5W0rSRWpAfpGd+XHJnNs10u2bupcZ3j6ClJxCtidlWx2KiIhIuc0JmcyPS+aOYa0J8fOyOpwGJcjXi4t7Nmf6xkQy8zWAnMiZUpIuUgMWbE8mv9iupu4VDIltDMCSnSkWRyIiInLUewt3EeTjyQ2DWlkdSoN0bf+WFBQ7mLYuwepQROotJekiNeDXTYdoHOBNfzcf1b2iZiF+tIkIYOkuJekiIuIadiVnM3PzYW4c3FJ30WtJt6gQurUI0QByImdBSbrIWSootrMgLplxXZri6aE/qYqGxoazYm8aRSUOq0MRERHhvYW78fG0ceuQ1laH0qBdOyCG7UnZrD2QbnUoIvWSMgpxDeu/MR/10PI9qeQV2RnXOdLqUFzOkNhw8orsGuVVREQsF5+Wx8/rE7mmfwyNA32sDqdBu6hHcwK8Pfjq+AHk6nF9T6QuKUkX15AZbz7qoXnbkvHz8mBQ28ZWh+JyBrZpjM2AJWryLiIiFvt06T4M4M7hbawOpcEL8PHk4l4t+HXjITLzKgwgV4/reyJ1SUm6yFlwOp3M25bEsHbh+Hp5WB2Oywnx86J7VKj6pYuIiKVyCkv4fnU853VrRrMQP6vDcQvX9o+hsMTBD2sPWh2KSL2jJF3kLGw9lEViZgFjO6mpe2WGxDZmfXwG2QWaikVERKwxdXU82YUl3DpUfdHrStcWIfSICuHrlRpATuR0KUkXOQvztiVjGDCqYxOrQ3FZQ2LDsTucrNybZnUoIiLihhwOJ5P/3EevmFB6RodaHY5buaZ/DLuSc1h7IMPqUETqFSXpImfh921J9IwOJSJIA9BUpndMGL5eNvVLFxERSyzYnsy+1Dxu0Yjude6CHs3x9/bgu1Xqhy5yOpSki5yhpKwCNh7MVFP3Kvh6edCvVSP1SxcREUt8unQfTYN9mdC1qdWhuJ1AH0/O79aMGRsTyS0ssTockXpDSbrIGZq3LRlASXo1DI0NZ0dSDslZBVaHIiIibmRvSi5LdqVw/cAYvDxU7bXCVf2iyS2y89umQ1aHIlJv6L+VyBmaty2JqDA/2kcGWh2KyxsSGw7A0t26my4iInXn21UH8LAZXNU32upQ3FbflmG0CQ/gu9Vq8i5SXUrSRc5AfpGdJbtSGNspEsMwrA7H5XVuFkyYvxdLdqZaHYqIiLiJohIHP6w5yOiOTWgS7Gt1OG7LMAyu7BvNqn3ppOUWWh2OSL2gJF3kDCzZlUJhiUNN3avJZjMY3DacpbtSNA2LiIjUiXnbkkjJKeKa/rqLbrXLe7fAw2awJTHL6lBE6gUl6SJnYN62JIJ8POnfupHVodQbQ2LDOZxVwO4juVaHIiIibuCbVfE0C/FlRHtNk2q1JsG+jOoQwZbELOwOXawXqYqSdJHT5HA4mReXzPAOEXh76k+ouoaW9kv/U/3SRUSklsWn5bF45xGu7BuNh03d0lzBlX2jyS0sYV+qLtaLVEUZhshp2piQyZHsQsZ20pX50xHT2J/oRn4s2akkXUREatf3pYOUXdU3yuJIpMzojk3w9/Fgc0Km1aGIuDwl6SKnad62JDxsBqM6KEk/XUNjw1m2J5USu8PqUEREpIEqsTv4bvVBhreLICrM3+pwpJSXh43OzULYcySXI9kaQE7kVJSki5ymuVuT6NMyjFB/b6tDqXcGtw0nu6CETbqKLiIiteSPnUc4nFWgAeNcUNcWwTicTqatS7A6FBGXpiRd5DQcTM8j7nC2mrqfocFtGwOoybuIiNSaH9Ym0CjAmzGagcXlNA7woXmoL9+tjtdsLyKnoCRd5DTMj0sG0NRrZ6hxoA9dmgezeJeSdBERqXlZBcX8vjWJC7s3w8tD1VxX1LV5CDuTc1gXn2F1KCIuS/+9RE7D3K1JtAkPoE1EoNWh1FvD2kWw7kA6uYUlVociIiINzKzNhykscXBJrxZWhyKVaN80CD8vD75ffdDqUERclpJ0kWrKLihm+Z5UxnbWXfSzMaxdOMV2Jyv2plodioiINDDT1iXQqrE/PaNDrQ5FKuHj6cF53ZoxfUMi+UV2q8MRcUlK0kWqafHOFIrtTsZ0VH/0s9GnZRg+njb+2KEm7yIiUnMOZxawbE8qF/dsgWFobnRXdlXfKHIKS5i5+ZDVoYi4JCXpItX0+7YkQvy86NMyzOpQ6jVfLw8GtGnMEvVLFxGRGvTLhgScTtTUvR7o37oRrRr7813pfPYiciwl6SLVYHc4WRCXzOiOTfDUQDRnbVhsOLuScziUmW91KCIi0kD8tC6RntGhtA4PsDoUqYJhGFzZN5rle9LYn5prdTgiLkfZhkg1rD2QTnpeMWM09VqNGNouHDC7EIiIiJyt7Yez2XYoi0t1F73euKx3C2wGTF2jAeREjqckXaQaft+WhJeHwfD2EVaH0iB0bBpEeKCP5ksXEZEaMW19Ah42gwu6N7M6FKmmZiF+DG8fwdQ1B7E7NGe6SEVK0kWq4fetSQxo3ZhgXy+rQ2kQDMNgWLtwlu5KwaGCWUREzoLD4eTndQkMbxdO40Afq8OR03BV32gOZRZonBqR4yhJF6nCvpRcdh/JVVP3GjasXTipuUVsPZRldSgiIlKPrdyXRmJmgQaMq4fGdGpCmL+XBpATOY6SdJEq/L4tCYCxnTQ/ek0aGmv2S9fVcxERORvT1iUQ4O3BuM5NrQ5FTpOPpweX9GrB3C1JpOcWWR2OiMvwtDoAkVqxewF8ccnR17fMgpaDqt4uLw3mvwBxv0J+GoS2xLd4OJ2aXEh0I39znZJC+H0SbP4B7MXQfjxMeAV8Q47u59dHYO3n8JcV0Kh1TZ7ZseJ+hSX/B4c3g6c3tBwCY56GJp1q53g7ZpvHO7QBDBs07wWjnjj5Z5uZAItehp2/Q+4R8/Np2g3GPAUt+tAk2JcOkUEs2ZnC3SPanrj92s9hxQeQshN8AqHtGBj7DIREHV1n+0z4/VnIOAARHeDclyFmwNH3E9bAh2Ng9BMw/NGa/zxEROTsTAo59fs3zYDWw8yf8zNgwYtm2ZeTDMHNKOl8GfM29WF81xj8vD1Ova/5L8LO2WaZUZAFPkFmudT/Duh88dH1jmyHXx+GxHUQEAFDH4Q+Nx19vzAH/tPbLGtv/PmMTrtaivLgj3/DpqmQcxgCm0K3K2DEY+DlV7PHykuDZe/C3j8gfR8UZEJIC2g9Akb+A4KOu1GREQ/znoPd88zPI7wdDLgLet9Y9bGWvWN+j6WeAZ7xAv5duuDi96DXda5V3xKpY0rSpeGxF8PMx05/u6I8mHwBJG85uix1J9ezk25NUoBR5rKlb8Hy92D4YxDcDGY8CDZPuPgd8/0jO2DNpzDoL7VbYGz4Fn666+jrknzY/hvsWwK3zYUmHWv2eOu/gWn3ABX6kO9bDJ9dCNdPhTYjjy4/sh0+PQ/yKtwlz0uBPQugy6XQog9gNnn/fPl+Cort+HpVqFwt+jcseKHCtoWw6TvYvxTuWGBWFtL2wHc3QpPOcP0P8OMd8M1EuH8d+IWa281+AkKiYdBfa/azEBGRuuFdOp1aUZ5ZrlQsozMO4Pnnm7xh74qzx49V72vLT5C68+jrggyzHNu3GK74BLpeDvYS+PZa8yLAxK/MxHX6/WYS2nKwud2S/4PcFBj/Uo2d5gmcTvj2Gtiz8OiyzAOw5A3z4sENP4Fh1NzxUnfD4teOXZa2x3zsmA33LAX/Ruby7CT4eBxkJx5dN2kz/PJXyEk6+4viZd+5q9S3RCyg5u7S8Cx7F1J2gJf/6W234v2jhf+YZ+DRPSQ2GwtAj+SfIX6V+d6OWebz4Pug763mFd0ds4/uZ86T4BcGwx45yxOpoDAbVn969HVRHsz8u/lzaEv420bzar6HNxRmwex/Vn/fxQXmVfq8tMrXsZfAnCcAJ4S1Mo93/3oIiQFHMcx4yKxQlPnpLjMp9w2FKyfD4wfg0d0w8WuI7FK+2tB24RSVOFi5t8KxMw7AolfMn1v0hYd3wKX/M19nJcDC0krR7vlgL4IeE807+R0vMFs/HFxtvr9lGhxYBudMAi/f6n8eIiJSdyZlHvt4KtW8YwzQONZssQWwZvLRMnrsJPhHAox+EoBhHpsZkr+g6mP1uRlunwePx8Oje8zXZTZNNZ/TdkPqLmg93Lz43Lv0DnpZOZ950Kxn9LkJIjuf+Xkf78iOY+sSm384mqD3uwMe22s+g3nBe/MP1d93Rrx5gaIqzXrAVV/APw7CX9ear8FMxtd9cXS9hf86mqBf+j+znG7Rt/S9V8zjncqg+0743lODzc/S7h0M7caZ61lR3xJxEUrSpWHJSjSbhgVEHC1Yq6usgPYOgiF/g4DGfO11eYX3vzef7aV9pjy8jz7bC82f9yw0m9KNfhJ8g8/4NModWA7T7oXXOsCMB44u3zXXvAMAZsEV1tKsTLQeURrHAshNPfW+D2+C3x6D1zvAD7eZFwIqk7wV8kr31+ki83iNWkOnC81labshYW1pzCvMq/xgVqS6XGoWrAHh0PF8iOpbvtsBrRvj7WE7tl/61p/NxB/Mq+NBkdDjagjvYC7b/CM4HGaLCQCP0hH3PUu/D3shlBTB789A9EDzzoiIiNQPcTPMpt0AfW45erd435Kj6/S/C3wCyex2W/kij81Tq9734PvMMsg3GAIaQ/87j75nK21cWl7Gl5UtPscu//1Zs9wf9eRpnthJFOXCui/h4/Hwbj/z4nKZsjoHwIi/m3exR/y9wvtVnG9Jkbm/Ly6Dt7rDyo9OvX5kF7hjIXS+yOwG0LgtDHv46Ptpe8xnhwM2lyb84R3M8jko0iyvwSy/t55mF4CEtTTO2grAqpBx4F16k6Uu61siLkbN3aVhmfMkFOXAhFfNO7LVVVwAR+LMn8Nagc2DYruDHw/4Un599vBG87nlEDPB3TYdQmPM/tYdzjMLrtlPQGQ36FWNPlmVyTkCG76GtV8cbZYXEHFsf7hDG47+3Dj22J93zQWnA5I2HdsEHcw+Zpummv23Dq03l/mEmH3I/BtXHlNJQdVxH94AUX1gf4WK1JHt8FZP8+JJozYw9AHzzncpP28P+rYK448dR/jneZ2qPreU7VCYCRn7IGYgYJhX1btdCbvmg6cvNO8NK/4L6fvhigqtD0RExPWtKf2/7ekLPa89urwk/4RV52w9zJVlL8rK6OrKSTbHPQEwPI6WsY3bmWXu/mVmeRz3q7m85WBznJNN38O4580k/0wdXANrPzMvOheVXiBv3vvohW84Whb6hEBghPlzYIT5ujCz8vNNjjPvem/45ujF9YiOx5S9J+V9ktaHJYVHfw5qbj6n7zWPDyeW0WVO97tYc7SsfuXIYL4p6wJX2/UtERemJF0ajr2LzeZf0QPMgn3hy9XftiDDTGzBvIIMrNqbxqECLyhrKZ17xHwe8XdI3mb2gQazOdi5L5uFYtJmuGk62EobqdiLj16NPxWHHXbNMwvtHbPNK9GevtD5EuhxDcSOBY8Kf655Fe6Sl8Z7ws+5Fe5O71tqxrf1ZyjOMyskseeYhXbHC6puDh7eHmxeZlzbfjEH2XE6zIKzPKbSJuuZCUeXrXj/6M9HtpnN4EsKj7ngMLRdOK/O2k5ydgFNgnyrf27R/eGc52D+8/ByjNkC4qL/mFfaF79mnluL3ub6JUVH77SLiIhrSt0NexaZP3e59GgfaDDv9O6eb/688gPofycly/579P1TddmqaMX/YGaFPtMe3nDxu2Y5C2Z5eNn/4Me74LVYc5DUgfeaCfQn55qtyPqXjgdjLwGbR/X6huelmWPJrPvCbJ0GEBwF/W+HHtdCRPvj1i8tCyuWfWWvCzOP1knAHLhty4/mBfiDpV3z/BubcVYsC09HcT4sedP82cMbul91bFzHx3ZMGV0htqoUZMEms+l+ZpP+rDvQlDlbk7ioR/PaqW+J1BNK0qVhsJeYg8UZHnDeazUymMrv25Lx8azYI6R0n/6N4KZfzObkjmIIamoWkAteNBPelkNh7tOwerJ5hbxpd7jwLWjes/KDbZxSOigb5kWGHteYFZSyAdCqzXnion1LYPL55s+RXc19d78KAk9j3ne/ULNZ4PJ3zVFf3+x24jplhWNZU3Uw+xJe+73ZdPHT88z+8gtfOiZJH9E+gldnbWfh9iNc1Te6mudW+l0Mud8cTTb7kFnZ8fA0R3q1l5jjCuyebzbpT91lVliGPmg2dxQREdezZjLl/+v73nrse/3vNJPQgkxzxO/fJ3FNxffPNEGzF8HPfzG7ZbUfby5rOxoe3g6Z8eZddW//o+OcTPzaHB/ll7+aXdI8vKHrpTDh3ye/G13muxvNAeq8AsxyuMdEaDX8aJJZbWVlYYV6zszHYP1XZiydLjT3327cmX8mJYVmvGVjAJz3WjUGZjtJGV0dG6dAcS4AQUPvpMVvfny/Ot5M0mujviVST6hPujQMcTPMK9PtzjFfH9pojjBaJm2PeYW+Mn5h5tVygMJsnE4nv29LYnTrClOcBIQfu01AY7PAAHOk1/x0swnc2snmiKRtRpgDqqTsgCk3mHdzK2NUGNm8IMtMZotPbNpXrmLT9Ip9yQtzToy34r4Ls819n6r/eWXGvWBO7xYaAx4+5qjqFQfdCW5hPvtVuPPRfaLZPK9pN3MQHjAT6gr95Ts3C6ZpsC8L4pJP49wqrOPpY3ZR8PA8OtLr0AfMysl3N5kVuqs+M6dom/ME7Pr99M9dRERqV0kRrP/a/Dmyq9laqqLQGLhlpnnH2yuAfK9QfrYPpjistJl1WRlUlQF3wjMZ5mCm5zxnLrMXmYl/RTabOf6Kt//RcU5aDzfHVvnxDvMC+IRXoOc1Zr/yP1499XHL6hj2IrOcL8gCR0nl65eVhceX12VlYcU6Sdm+HSVH920/RZ3jVEoKYcr1sHOO+Xr8v47tbnc69Y/qWDO5dL/h2DpfzBV9oliyK4WD6XkV9leD9S2RekJJujQMReZVWHbMgg+GmY8KfZz4+V745f7Kt/f0MftsAaTvY9fhTA6k5XFhiwqFRNPuJ9+2bKTXAXeZ/a7LmuMNfxS6X2n2C888cOy0L8frcTXcOht6Xm/2pZ/7NPxfZ/jiUtj4nTmae0VlI66CeZf4+J8Nm9lXC8yRz+9eCgPuNhP0Ra+Y87t+PM4cMT4/o/K4KrLZzEFkHtgETyXDvcvM0dvLjhdTOld6s0o+p4rKBuIBDMNgVMcIFu9MoajEUfW5+YRAaKuT73fOkxDUDAb/1WzyV5hlDhzX+WIYWNpSYXc1RgAWEZG6tfXno9N2Hn8XvUxkF7j+B5z/TOAiv8/5udn9eOUcMt9rNaT6xzIMM5Ec8rejc26f6kL+iv+aI5aPf8lMgA+uMi8k9LutfIT58rK/Mtd8Y87/HdUXtv8K390Ar7c3Z0cpm5WkorKysDDT7BsP5nNZf/CKdZLz34ArPzNbAOxbDNPuhn+3g5/uNgdYcziq/EgAM0H/9rrSBN0w76CXDQhXJqy1WQ7Dycvo42M7lQMrzGbrAL2uB09vrugThdMJP6xJOHH9mqhvidQTStLFPU0KMR8/3XN0WbcrzOeibFLnvEYYWYxK+arC+1dyUr8/a87pObx0bvayO9dlzczKRoy1VdG7JGYgXPIuPLIDLnzbnEt893zziv1r7c1R3svEnnM0QV79iTlI2p6FsLe0L1+bUcfebW7a1bzi//B2cy7YNqPMSsaMB8x9f3eTWfE4ld0LzDsHBVlm37rVn5jzl4LZvC6k9C5Gu3HgUzrS6sZvzUrF4U2w9w9zWbOe4BNo/vzTPTAphJc2DiOnsITV+9LMfvi20s9u2bvmfKwbvzMHjQPoetnJmweWjfQ6dhJ4+VX4Hso+/7Lvw+PEbUVExFplF9a9A4/2f67I4TD7k6fvZ1t8Mh5HtvJS8b/NptKGB/S7/ei6C146Ws6n7zeXHVgBi16Fw5vNC/t5afDnO2ZrKzBbZJ1Mbqo5zknvG8yy1OYBGCcpW6oo470DoNd1cOssuG+NeYHAwxtWfwwfjYH/9Dl2xPaKdY5Fr5jxlk1PCkfrLGCOudLlErj+B3hgM4x6wrwIseEb+PxieLMrLK/Qf/9kSgrN+eF3zTUvvF/0H3P8mePZbGbzfjDL5Y3fmeX0snePfh5dLjm6/snqW2XKb6YY0PcWAKIb+TMktjHfr4nH4TiuC19N1bdE6gH9FkvD0Os681HRgpdgUengcbfMMu8on8rAe83BS5K3MHDvf1jnC5RdGO59I0T3O3GbspFeL3jj6BQgHSbA1mmw8n9mv/K9i8xRTxu1rd65+ASaTcv63FRhlNZvzf5ml5Qmxd7+ZtL9012Qsd+cXqV8+2CzedrJePqYd5a7Xm7esV/3lbnfrdPMZn+nmsZk3xKzonK8Rm3Mvnjlxw+C8S+a/fUS15kD75SxeZnN5k/C28PG/LhkBl/Q2RwsZsELkLDavNNQJqg5jPzHiRuXjfQa1f9oxSW6n9n0ftt0c9q4siZ17c+t/BxFRKTuHdkO+5eaP3e78sTB0sAcrHTmozDzUToDs3yA9NL3znnW7FZ1KrlHzL7MC1488T3DBqP+efLtFr4ETufRKde8A6D1MNj/J2yfCYnrzeWnU7aEx5pl7uinzYvLa78wk+Nd846WYV0vN8v/PQth1Yfmo0ybUZVPLxrSAkY8Zt5d3rvI3HfcDLMsHHh35THFrzzaHczpgF/uMx9lWg6FW0pHuh/5T9gxx5wr/cfjEvmRf4eQqKo/g/z0o3O3tx19zEWSq/pG87dv17N8TyqDY0ubztdGfUvEhSlJFynj5Qc3z6Bg9rNkrZ9GI1suno1amgn6oEoGG5v9hNk3u+Kc7N2vNgeVWT3ZvMLcorfZZMzjDP7cmnQ0E96xk2D7b8e+12OiWZFZ/AYkbTGvyLccbPYbb9Kx6n2HxsCof5gJ8Z4FVQ9SF9UXovpByk5zhPigZuYd9GEPHzsCL5ifmW+IOTJs8lYzOY/uZybYx/czLDWgTSPmb0/myQs6w4hHzXlXV3xgHs87AGLHmIPBBUWeuPG6L8zP4PZ5R5f5hcE138Lsf5p3EoKamncGWg6u+rMREZG6s7pC97TKmrobNuh6Oc74lRRlHsZueOPfqg8M+iu0H1f1MSI6QrerzIu/2Ulmn+2ACLNsG3jvyS/kl41zMubpo9OgAVz6Afz2qHmh3NMPBv4FhjxwWqcMmPWCjuebj+zDxzYZNwyY+I3Z133TD+Z4LkHNoNvl5p3kqgbINQyz+XebkWa3toSTNKk/U0GRcNscmPeseWGhKBfC25nN0HtXc0q09d8cnd71uO98fJemBPl68t3q+KNJem3Xt0RcjOF0nmQ06KNO+aZIjVlUOuDKiMesjQP4fnU8j07dyIy/DqVrixCrw3Ebk5fuZdL0rSx6dCQtGwdYHY6Iq6vu8Mkqx6VBWbIzhes/XsF71/XmvG7NrA5HTlc163tPTdvMlNXxrPjHGMICNIWqNEinLMfVJ13kOHO3JtEsxJcuzU/R9Ftq3OiO5h3y+WWjvIuIiBznp3UJBPl4MrrjaUwjKvXOtQNiKCpxMHXNQatDEbGEknSRCgqK7SzemcLYTpEYNTDXulRfTGN/2kYEKEkXEZGTyi+yM2vzIc7r1gxfLw0C2pB1ahZMv1ZhfLli/4kDyIm4ASXpIhUs2ZlCfrGdcV1O0u9Zat3ojk1YsSeN3MJTzB0rIiJuae62JHKL7FzSq5pzoku9dv3AluxPzWPxrhSrQxGpc0rSRSqYuzWJIB9PBrRuXPXKUuNGdWxCkd3BEhXIIiJynJ/XJdAsxJcBrRtVvbLUe+d2bUrjAG++XL7f6lBE6pySdJFSdoeTeXFJjOzYBG9P/WlYoV+rRgT5eLJATd5FRKSC1JxCFu04wkU9m2OzqTuaO/Dx9ODqftHM25ZEQka+1eGI1CllIiKl1senk5JTxDmd1dTdKl4eNoa1D2fB9mSqmHlCRETcyK+bDlHicHKpmrq7lWv6x+AEvllxwOpQROqUknSRUnO2JuHlYTCyQ0TVK0utGdMxkqSsQjYezLQ6FBERcRE/rUugY9MgOjbVzCvuJLqRP6M7NOHbVfEUlTisDkekzihJFyk1d2sSA9s0JtjXy+pQ3NqYTk3wsBnM3nLY6lBERMQF7EvJZd2BDN1Fd1PXD2pJSk4hMzcfsjoUkTqjJF0E2H0khz1HctXU3QWE+nszsE0jJekiIgLAtPUJGAZc1LO51aGIBUa0i6BVY38+XbrP6lBE6oySdBFgzpYkAMZ2UpLuCs7t0pTdR3LZlZxtdSgiImIhp9PJtHUJDGzdmGYhflaHIxaw2QxuGdKa9fEZrNmfbnU4InVCSboIMGNjIj2jQ2keqgqAKzinc1MAZpdePBEREfe0Pj6Dfal5auru5q7oE0WQryefLN1rdSgidUJJuri9vSm5bEnM4oLuzawORUo1DfGlZ3QoszarybuIiDubti4Bb08b53ZranUoYqEAH0+u6R/DrM2HNR2buAUl6eL2ZmxIBOC8bkrSXcm5XZuyKSFThbGIiJsqtjuYsfEQ53SK1KCuwk2DWwHw2Z/7LI1DpC4oSRe3N2PjIfq2DFNTdxdzbhfzrsnMTRrNVUTEHS3eeYTU3CIu1oBxArQI9ePcrk35ZuUBcgtLrA5HpFYpSRe3tjMpm+1J2Wrq7oJahQfQtUUw0zcqSRcRcUc/rEmgUYA3Izs0sToUcRG3DmlNdkEJU9cctDoUkVqlJF3c2oyNhzAMNXV3VRd2b86G+Az2p+ZaHYqIiNShjLwi5m5N4uKezfH2VHVVTH1ahtEzOpRPl+7F4XBaHY5IrdF/PXFbTqeTGRsTGdC6EU2Cfa0OR07igh5mE8cZupsuIuJWpm9IpMju4PLeUVaHIi7mtqGt2Zeax/y4ZKtDEak1StLFbcUdzmb3kVwu6K6+bq6qRagffVuG8cv6RKtDERGROjR1bQIdmwbRpXmw1aGIizm3a1Oah/jy8RJNxyYNl5J0cVszNibiYTOY0FXTuriyi3o2Z3tSNtsPZ1sdioiI1IGdSdlsiM/gij5RGIZhdTjiYrw8bNw8pBXL9qSyOSHT6nBEaoWSdHFLZlP3Qwxu25jGgT5WhyOnMKFrM2yG2fRRREQavqlrD+JhM7i4ZwurQxEXNbF/DIE+nny4eI/VoYjUCiXp4pY2J2SxPzVPo7rXAxFBPgyJDeeXDYk4nRokRkSkIbM7nExbl8CoDhFEBOkiupxcsK8XE/tFM2PjIRIz8q0OR6TGKUkXtzRjYyKeNoPxXdTUvT64sHtzDqTlsfGgmrWJiDRki3ceISmrkCv6aMA4ObVbhrYGYPKf+6wNRKQWKEkXt1PW1H1ou3BC/b2tDkeqYXzXpnh72PhpXYLVoYiISC2auuYgYf5ejO4YaXUo4uJahPpxXrdmfLPiANkFxVaHI1KjlKSL21kXn0FCRr5Gda9HQvy8OKdzJL9sSKSoxGF1OCIiUgsy84uZszWJi3pobnSpnjuGtSa7sIQpq+KtDkWkRuk/oLidGRsO4e1hY1wXXaWvT67oE0VabhHz45KsDkVERGrBjI3mhdgr+kRbHYrUE92jQhnQuhGfLNlLsV0X8aXhUJIubsXhcPLbpkMMbx9BsK+X1eHIaRjWLpwmQT5MXXPQ6lBERKQWTF1zkA6RQXRtobnRpfruHN6GxMwCftt0yOpQRGqMknRxK6v3p3M4q4ALe2hU9/rG08PGZb2jWLD9CMnZBVaHIyIiNWj3kRzWHdDc6HL6RnVoQpuIAD5cvEezwEiDoSRd3MqMjYn4eNoY00lN3eujK/pElU/PIyIiDccPa0rnRu+l8WLk9NhsBrcPbcPmhCyW70mzOhyRGqEkXdyG3eHkt02HGd2xCYE+nlaHI2cgtkkgvWJC+X71QV0tFxFpIOwOJz+uTWBE+wiaBPlaHY7UQ5f1bkHjAG8+WrzH6lBEaoSSdHEbK/akkpJTqFHd67kr+0SzMzlHc6aLiDQQf+5O4XBWgeZGlzPm6+XBDYNaMi8umV3J2VaHI3LWlKSL25i+8RB+Xh6M6hhhdShyFi7o0QxfLxtTVmu6FRGRhmDKqnhC/b0Y06mJ1aFIPXbDwJb4eNr4eMleq0MROWtK0sUtFNsdzNp8iDGdmuDvrabu9VmwrxfndWvGL+sTyS0ssTocERE5C2m5RczZksRlvaLw8fSwOhypxxoH+nB5nyh+WJtASk6h1eGInBUl6eIWlu1OJT2vWE3dG4hr+8eQU1jC9A2JVociIiJn4ce1BymyO7i6n+ZGl7N329DWFJU4+HzZfqtDETkrStLFLfy26RAB3h6M7KCm7g1Bn5ZhtI8M5OuVB6wORUREzpDT6WTKqnh6xYTSoWmQ1eFIA9A2IpCxnSL5cvl+CortVocjcsaUpEuDV2J3MHvLYcZ0isTXS03pGgLDMLi2fwwbD2ayOUEDyImI1EdrD2SwMzmHibqLLjXojmGtScst4oe1B60OReSMKUmXBm/5njTS84o5r1tTq0ORGnRp7yh8PG26my4iUk9NWXWAAG8PdUWTGtW/dSN6RIXw8eK9OByarlXqJyXp0uD9uukQ/t4ejOygUWMbkhA/Ly7o3pyf1yWQowHkRETqleyCYqZvOMSFPZoT4KMBXaXmGIbB7cPasCcll3lxyVaHI3JGlKRLg1ZidzBny2FGd2yipu4N0LUDYsgtsvPLeg0gJyJSn8zYeIj8YrsGjJNaMaFrU1qE+vHhH3usDkXkjChJlwZt5d40UnOLOL9bM6tDkVrQOyaUjk2D+HqlRnEVEalPvl0VT4fIIHpGh1odijRAnh42bh3ampX70tgQn2F1OCKnTUm6NGi/bjqEn5eaujdUhmFw7YAYNidksemgBpATEakPth3KYkN8BhP7R2MYhtXhSAN1db9ognw9+XCx7qZL/aMkXRosu8PJ7NKm7n7eaureUF3cswW+XjbdTRcRqSemrIrH29PGpb1aWB2KNGCBPp5cOyCG3zYdIiEj3+pwRE6LknRpsFbuTSMlp4jz1NS9QQvx8+LC7s35eX0i2QXFVocjIiKnUFBs58e1Bzm3S1NC/b2tDkcauBsGtgTgmxWaCUbqFyXp0mDN2XoYb08bozpGWB2K1LJrB8SQV2Tnlw0aQE5ExJXN3nKYrIISzY0udSIqzJ/RHZvw7aoDFJU4rA5HpNqUpEuD5HQ6+X1bEkNjw/H31tQuDV3PaHMAuW9XxlsdioiInMJXKw4Q08ifgW0aWx2KuInrB7YkJaeI2VsOWx2KSLUpSZcGaWdyDvFp+YzppAHj3IFhGEzsF82mhEw2J2gAORERVxR3OIuVe9O4fmAMNpsGjJO6MbxdBDGN/PliucaukfpDSbo0SL9vSwJgTMdIiyORunJpryh8PG18u0r9zkREXNGXy/fj7Wnjyj5q6i51x2YzuG5ADCv3prH9cLbV4YhUi5J0aZB+35pEtxYhNA3xtToUqSMh/l6c160ZP69LJL/IbnU4IiJSQXZBMT+tTeDC7s0JC9CAcVK3ruwbjbenja9W6G661A9K0qXBSckpZF18BmM76S66u7m6XzTZhSX8uumQ1aGIiEgFP61LILfIzo2DWlodirihRgHeXNCtGT+uTSCnsMTqcESqpCRdGpz5cck4nag/uhsa0LoRbcID+HalmryLiLgKp9PJF8v20z0qhB7RoVaHI27quoEtySksYdq6BKtDEamSknRpcH7fmkSzEF+6NA+2OhSpY4ZhcHW/aFbvT2dnkvqdiYi4ghV709iZnMP1A3UXXazTOyaUTs2CNXaN1AtK0qVBKSi2s3hnCmM7RWIYGjnWHV3eJwovD4NvV2k6NhERV/DFsv2E+HlxUY/mVocibqxsJpjNCVmaCUZcnpJ0aVCW7U4lv9iupu5uLDzQh3M6R/Lj2oMUlmgAORERKx3OLGD2lsNc2ScKXy8Pq8MRN3dJzxb4eNqYogv54uKUpEuDMndbEgHeHgxq29jqUMRCE/vFkJ5XzOwtSVaHIiLi1j5btg+H08lNg1tZHYpI+Uww09YnaCYYcWlK0qXBcDqdzNuWxLB2Efh46mq9OxsaG05UmJ8GkBMRsVBeUQlfrzjAuM5NiW7kb3U4IkDpTDAFJczcrJlgxHUpSZcGY3NCFklZhYztrKnX3J3NZnB132j+3J3K/tRcq8MREXFLP6xNIDO/mNuHtbY6FJFyA1o3olVjf41dIy5NSbo0GL9vS8IwYFSHCKtDERdwRd8obAYqhEVELOBwOPlkyV56RIXQp2WY1eGIlDNngolh5d409hzJsTockZNSki4Nxu/bkugTE0bjQB+rQxEX0CzEj1EdmvDDmoOU2B1WhyMi4lYWbE9mb0outw1ro9lWxOVc3qcFHjaDKat1IV9ck5J0aRAOZeazJTGLMZ3U1F2OuqpfNMnZhSzcfsTqUERE3MrHS/bSLMSXCV2bWh2KyAmaBPkypqN5Ib9YF/LFBSlJlwZh3rZkAMZq6jWpYHTHJoQH+qjJu4hIHdqamMWfu1O5aXArvDxU1RTXNLF/NCk5ReV1SBFXov+c0iDMj0smppE/sU0CrQ5FXIiXh43L+7RgwfZkkrMKrA5HRMQtfLh4D/7eHlzTL8bqUEQqNbxdBE2DfZmySjPBiOtRki71Xn6RnaW7UhjdsYn6vckJruobjd3h5Ie1CVaHIiLS4B1IzeOXDYlcNyCGEH8vq8MRqZSnh40r+0axaMcREjPyrQ5H5BhK0qXeW7orhcISB2PVH11Oom1EIP1ahfH96nicTqfV4YiINGjvL9qNh83gjmFtrA5FpEpX9Y3GCXy/+qDVoYgcQ0m61Hvz4pII8Pagf+tGVociLuqqvtHsScll1b50q0MREWmwDmXmM3VNPFf1jaJJsK/V4YhUKbqRP0Njw/ludTx2hy7ki+tQki71mtPpZN62ZIa3j8DbU7/OcnLnd29GoI8nUzSAnIhIrfnfH3twOOGu4W2tDkWk2q7uF01CRj5Ld6VYHYpIOWU1Uq9tTsgiObtQU6/JKfl7e3Jhj2b8tukQ2QXFVocjItLgpOQU8s3KA1zSswXRjfytDkek2s7pHEmYvxffagA5cSFK0qVemxeXhGHAyA4RVociLu6qvtHkF9uZvuGQ1aGIiDQ4nyzZS2GJg3tH6S661C8+nh5c1juKuVuTSM0ptDocEUBJutRz8+OS6RUdSnigj9WhiIvrGR1K+8hApqxWk3cRkZqUkVfEF8v2c17XZrSN0FSoUv9c3S+aYruTn9ZpJhhxDUrSpd5Kyipg48FMNXWXajEMg6v7xbAhPoO4w1lWhyMi0mB88McecopK+OuYWKtDETkj7SOD6BUTypRVmglGXIOSdKm3FsQlAzCmUxOLI5H64tJeLfDyMDSAnIhIDUnOLmDy0n1c2L05HZsGWx2OyBmb2C+anck5rD2QYXUoIkrSpf76fVsyLUL96BAZZHUo1TJt2jS6d++Oj48PrVu35o033jjl+g8++CCGYfDII48cszwuLo4BAwYQEhLCxIkTycnJOeb9P/74gxYtWpyw/GQmT56MYRgnXXfSpEmEh4eXv963bx+GYZQ/AgICaNu2Lddddx2LFy8+Yfubb76Zvn37VhlDXWoU4M24zk35aV0ChSV2q8MREan33l+4myK7gwfGtrM6lFqncrxhu6B7cwK8PZiiAeTEBShJl3opv8jO0l0pjOnUBMMwrA6nSkuXLuWyyy6jf//+TJ8+nVtvvZW///3vvPnmmyddf+vWrXz88ccEB594V+Lmm28mNjaW7777jq1bt/Kvf/2r/D2Hw8Hf/vY3XnrpJQIDa6df4GuvvcayZcv47bffeOqpp0hNTWX48OE8++yztXK8mnZVv2gy8oqZuzXJ6lBEROq1xIx8vlp+gMt7t6BNA++LrnK84Qvw8eTCHs2ZvkEzwYj1lKRLvbRoRzL5xXbGd2lqdSjV8txzzzFkyBA++ugjxo0bx1NPPcX999/Pc889R1FR0Qnr//Wvf+Vvf/sbYWFhxyzPyclhxYoVvPnmm4wfP54nnniCuXPnlr//ySef4OXlxQ033FBr59KhQwcGDhzIiBEjuPnmm5k1axZPPfUUkyZNYuHChbV23JoyNDac5iG+avIuInKW/jN/J06c3D+m4d9FVznuHq7uZ84EM2OjZoIRaylJl3pp5ubDhPl7MaB1I6tDqZb169dzzjnnHLNs3LhxpKens2zZsmOWT506lbi4OB5//PET9lNWEfDz8wPA39+/fFlWVhZPPvkkb731Vp23LnjmmWdo3rw5//3vf+v0uGfCw2ZwRd9oluxK4WB6ntXhiIjUS/tScvlu9UGu6R9DVFjDnxdd5bh7KJsJ5ltdyBeLKUmXeqewxM68bcmM69wUT4/68StcUFCAt7f3McvKXm/btq18WX5+Pg8//DAvv/wyAQEBJ+ynUaNGtGrViv/85z+kpaXxv//9r7y/2PPPP8/YsWMZNGjQacdnt9spKSk55uFwOKq9vYeHB6NHj2b58uWnfWwrXNknCoCpaw5aHImISP30+twdeHkY3DfKPUZ0VznuHirOBLPtkGaCEet4Wh2AyOlasjOFnMISzu1WP5q6A8TGxrJq1apjlq1cuRKAtLS08mUvvfQSzZo14/rrr690X++99x5XXnkl//znP2nXrh3vvvsuu3bt4qOPPmLTpk1nFF9oaOhJlzdu3Lja+4iKiiIpqX70845u5M+QtuF8v/og949uh83m+uMaiIi4inUH0pm+IZG/jo6lSbCv1eHUCZXj7uPSXi14ZWYcU1bFM+miLlaHI26qftyGFKlg5ubDBPl6MqRteNUru4i7776badOm8eGHH5Kens7s2bPLR4W12cw/w7179/Laa69V2cxtwoQJJCcns337drZt20ZMTAwPPfQQDz74IFFRUbz77rvExMQQExPDe++9V634/vjjD1atWnXM44477jitc6xv84pe3S+ahIx8lu5OsToUEZF6w+l08uKv2wgP9OGuEW2tDqfOqBx3H40CvBnXJZKf1iVQUKyZYMQaupMu9Uqx3cHcrUmc0ykSb8/6c43p1ltvZcOGDdxzzz3ceeed+Pv788orr/DXv/6Vpk3NFgGPP/44EyZMoEOHDmRkZADmKK+FhYVkZGQQEhJSXuj7+/vTvn17AObOncuGDRuYMmUKGzZs4KmnnuLPP/8EYNCgQQwdOpTu3bufMr5evXqdMIrsjBkzTuscExISiIyMPK1trDSuSySh/l58uyqeYe0irA5HRKRemL3lMKv3p/OvS7sR6OM+1UiV4+5lYr8YZmw8xOwth7m4ZwurwxE3VH+yHBHgz92pZOYXc27X+tPUHcy+Xu+88w5Hjhxh48aNJCUlMXDgQIDy5+3bt/Pjjz8SFhZW/oiPj+edd94hLCyMhISEE/Zrt9t58MEHefXVV/Hz82PhwoWMHj2ajh070rFjR8aMGcOiRYtq/fxKSkqYP3/+GfWjs4qPpweX9GzB3C1JpOeeODKviIgcq6jEwcsz42gfGchVfaOsDqdOqRx3L4PbNiYqzI9vV2oAObGG+1wClQbhp7UHCfb1ZESH+nnns6zQBrNP2uDBg+nYsSMAH330ETk5OcesP3HiREaMGME999xDRMSJ5/z+++8TFhbG1VdfXb4sL+/oiOW5ubl10nztueeeIzExkbvvvrvWj1WTJvaPZvKf+/hudbxbNdsUETkTXy7fz77UPD69pV+9Gbi1pqkcdw82m8HEftG8NmcHu5JziG1SO3PWi1RGSbrUGzmFJczeksSlvVvg4+lhdTinZfny5SxZsoSePXuSlZXFN998w+zZs1myZEn5OmWju1bk6+tLdHQ0I0eOPOG9tLQ0nn32WWbPnl2+bPjw4Tz22GN88sknOJ1O5s+fz8svv1yj57J9+3bCw8MpKipi7969fPvtt8yaNYtJkyYxYsSIGj1WbevYNJiBbRrx+bL93Da0tdtWOkVEqpKZV8zb83cyrF04I9vXzwvlZ0PluPu5ul8Mb83byZfL92sAOalzStKl3pi1+TD5xXYu713/+gZ5eXkxZcoUJk2ahM1mY9iwYSxdupRu3bqd8T4nTZrERRddRO/evcuX9erVi1dffZUnnngCgNdee40ePXqcdfwVPfLII4BZ8WjWrBmDBg3ijz/+YNiwYTV6nLpy65DW3PnFGmZtOcwF3ZtbHY6IiEt6d+EuMvOL+ceETnU+h7crUDnufiKCfDivWzN+WHOQR8d3IMCNxmAQ6xlVNKHRMI9SNxa9aj6PeKzSVa79cDkJGfksfGSkW1YQpHbYHU5Gv76QRgHe/HTvEKvDEamu6v4TVDkuZy0+LY8xry/i4p7N+feVNZswipupRn3PlazZn8bl7y/jhUu6cv3AllaHIw3LKctxte2UeiExI59le1K5tFcLJehSozxsBrcMbsW6AxmsPZBudTgiIi7n5VlxeNgMHh7XwepQROpU75gwOjcL5otl+zVFndQpJelSL0xbn4DTCZf1cq/RZKVuXNE3miBfTz78Y4/VoYiIuJTle1L5deMh7h7RlqYhvlaHI1KnDMPgxkEt2Z6Uzcq9aVaHI25ESbq4PLvDybcr4+nfuhExjf2tDkcaoEAfT24c1JJZWw6zMynb6nBERFxCid3BpF+20CLUj7tGtLE6HBFLXNyzBcG+nny+fL/VoYgbUZIuLm/RjmQOpOVx4yD1BZLac9vQNvh5efDOgl1WhyIi4hK+WRVP3OFsnjy/E75e9WtWFZGa4uftwZV9o5m9+TDJWQVWhyNuQkm6uLzP/txPkyAfxndpanUo0oA1CvDmhkEtmb4hkd1HcqreQESkAcvIK+L1OdsZ1KYx53ZV+Svu7YaBLbE7nXy+THfTpW4oSReXtjcll0U7jnDtgBi8NIe11LI7hrXB29PGu7qbLiJu7o25O8guKOGZizprwFZxe63CAzinUyRfLN9PbmGJ1eGIG1DWIy7ty+X78bQZXNs/xupQxA2EB/pw/YCWTFuXwK5k9U0XEfe07VAWXy7fz/UDYujYNNjqcERcwl0j2pCZX8x3q+OtDkXcgJJ0cVnZBeY/wnO7NqVJsEaUlbpx76hYArw9eXlmnNWhiIjUOafTybPTtxDi58WD57S3OhwRl9GnZSP6tAzj4yV7KbE7rA5HGjgl6eKyvlx+gOyCEu4YphFlpe40CvDmnlFt+X1bMsv3pFodjohInZq5+TDL96Tx8LgOhPp7Wx2OiEu5c3gbDqbnM3PzYatDkQZOSbq4pIJiOx8v2cOwduH0iA61OhxxM7cOaU3zEF/+9ds2HA6n1eGIiNSJ/CI7L/66jU7NgrlG3cxETnBOp0jahAfwwR+7cTpVP5DaoyRdXNJ3q+NJySniL6NirQ5F3JCvlwcPj+vAxoOZTFufYHU4IiJ14oM/dpOQkc+kCzvjYdNgcSLHs9kM7h7Rls0JWSzYnmx1ONKAKUkXl1Nsd/DBoj30bRnGgNaNrA5H3NSlvVrQMzqUf/22jcz8YqvDERGpVQfT83h/4W4u6N6MAW0aWx2OiMu6tHcLYhr5839zd+puutQaJenicn5Yc5CEjHz+MipW076IZWw2gxcu6UpabhGvzd5udTgiIrXqpd/iMAz453mdrA5FxKV5edi4b3QsmxIy+X2b7qZL7VCSLi6loNjOW/N20jM6lJEdIqwOR9xc1xYh3DioFV+u2M/GgxlWhyMiUiuW7U7l102HuHdkLM1D/awOR8TlXdarBS0b+/Pm7zt0N11qhWdN7qzYXsxba99i/ZH1bEvdRpGjCIDPzv2M3pG9a/JQ0kB9teIAhzILeP3KHrqLLuUOZB3gk82fsD55PXsy9+DEiYfhwfob19f6sR8a155fNx3inz9tYtq9Q/D00LVNEWk4iu0Onv55M1Fhftw5vGZnU1G9UBoqTw8bfx3djke+38CcrUmM79LU6pBcmpX1uPqqRmub+fZ8Ptv6GRuObCj/RyxSXYUlDt5bsIuhseEMjg23OhxxITszdvLDzh/YnbkbJ3V7xTrY14tnL+rC5oQsPl6yt06PLSJS2z5ZspedyTlMurALvl4eNbpv1QulIbukZ3PahAfw6qw4ijVv+ilZWY+rr2o0Sfc0PJnYYSIvD3uZq9pfVZO7Fjew7kA6qblFPDK+g9WhiIuJ9I/kjm538M7od+gW3q3Ojz+ha1PGdY7kjbk72JeSW+fHFxGpDYkZ+bw1bydjO0UytnNkje9f9UJpyDw9bPzjvE7sPpLL1ysOWB2OS7O6Hlcf1WiS7u/lzxMDn+D8NufT2E8jg0r15RfZWb0vjXGdI+mpedHlOF3Du3J/7/sZET0CHw+fOj++YRg8d3FXvD1s/POnTep/JiINwvMztuJwOnnmws61sn/VC6WhG9upCYPbNubN33eQmaeZYCpjdT2uPlLnSnEJq/anUWR38PA43UUX19Q0xJd/nNeJP3en8v3qg1aHIyJyVhZuT2bm5sP8dXQ7ohv5Wx2OSL1kGAZPnN+JjPxi3lmw0+pwpAFRki6WS8oqYN2BdDo2DaZD0yCrwxGp1MR+0fRv3YgXft1KclaB1eGIiJyRgmI7z/yyhTbhAdw+rLXV4YjUa12ah3BVn2gm/7mP3UdyrA5HGggl6WK5d+bvwuGAwW3VFE5cm81m8PJl3SgocTBp+harwxEROSMfLNrD/tQ8nru4Kz6eNTtYnIg7emR8B/y8PPjHj5twONQlTs7eGSXpqw6vottn3Y55TNs1rYZDE3dwIDWPb1YeoFtUCKH+3laHI1KlNhGB/G1MO37bdJjZWw5bHY6IyGnZn5rLuwt3cUH3ZgxtVzMzqaheKO4uIsiHJ87vxMq9aUxZHW91ONIA6E66WOrN33fgYTMY2EZ30aX+uHN4Gzo1C+apaZvJzNdAMSJSPzidTp7+eQteNoMnz6+dweJE3NVVfaMZ2KYR//ptm7rEyVnzPJON+jXtx6abNp30vfSCdAAKSo7+cuYU55BekI6Phw/+XhqcREw7krL5aX0Cdw5rQ6DPyX+fRACKHcXkFJn9vEocJeXLy/7fBHgF4O1Rdy0xvDxsvHJ5Ny55dykvz4zjpcs0nYiIuL5fNiSyaMcRnrmwM01DfGtsv6oXipiDyL10WXfGv/kHT07bzAc39MEwDKvDcgmuVo+rD4wqphI67U4V3T6rvLJ6UduLeHHoi6e7S2mg7vpiNX/uSuWPx0YRtvpNc+GIxyyNSVzTqsOruHX2rZW+//yQ57kk9pK6C6jUv37bxv/+2MO3dw5UaxCpa9Wt+alzpACQllvE2DcWEdPInx/uGYyHrW6SB9UL5RiLXjWfG3B978M/9vDib9t46bJuXNM/xupwXIKr1uMsdsp/wmruLpbYEJ/B7C1J3DG8DWEBunIm9dODY9sT08iff/y4iYJiu9XhiIhU6oUZW8nKL+aVy7vXWYIu4o5uG9qaobHhPDt9C7uSNdq7nJkav5MuUh3Xf7SCrYey+OOxUQT6eLrFlVVpmJbuSuG6j1Zw78i2PHZuR6vDEfehO+lSbX/sOMKNn6zk/tGxPDSug9XhiDtzk/peclYB5761mKbBvvz0l8GaRUFORnfSxbX8uSuFJbtSuHdkWzNBF6nHhsSGc2WfKD74Yw9bEjOtDkdE5Bi5hSX886dNtI0I4C+jY60OR8QtNAn25dXLu7P1UBbPTt9qdThSDylJlzrldDr595ztNAvx5fqBLa0OR6RGPHF+J8L8vXn8h02U2B1WhyMiUu7F37aRkJHPy5d31908kTo0tnMkd41ow9crDvDtygNWhyP1jJJ0qVPztiWz7kAG949ph6+XKgvSMIT6e/PsRV3YlJDJJ0v3Wh2OiAgAC+KS+XrFAe4c1oZ+rRpZHY6I23lsfEeGtQvn6Z+3sO5AutXhSD2iJF3qjMPh5LU522kdHsAVfaKsDkekRp3XrSnndI7kjbk72J+aa3U4IuLm0nKLeHTqRjo2DeKhce2tDkfELXnYDP5zTS8iQ3y464s1JGbkWx2S1BNK0qXOTN+YSNzhbB48pz1eHvrVk4bFMAyev7grXjYb//xpE1UMyikiUmucTif//HETWfnF/N/VPdXMXcRCof7efHRjP/KL7Nz86Uoy84utDknqAWVKUieK7Q7emLuDjk2DuKBbM6vDEakVTUN8eezcDizdlcq09QlWhyMibuqHtQnM2nKYh8e1p1OzYKvDEXF7HZoG8cENfdibksudn6+msETTtsqpKUmXOvH96oPsT83j0fEdsGl+VmnArhvQkl4xoTw/YxvpuUVWhyMibmZHUjZPTdvMgNaNuH1YG6vDEZFSg2PDee3KHqzYm8Yj32/E4VCLO6mcknSpdflFdt6et5PeMaGM7tjE6nBEapXNZvDSZd3Iyi/mX79tszocEXEjuYUl3PvVWgJ8PPnPNb3w0EVxEZdycc8W/P3cjkzfkMhzM7aqa5xUSkm61LqPFu/hcFYBj0/ohGGowiANX8emwdw+rA3frznIst2pVocjIm7A6XTyjx83sedIDm9f05Mmwb5WhyQiJ3H3iDbcPrQ1k//cx8sz45Soy0kpSZdalZxdwPuLdjO+SyT9W2v6F3EffxvTjuhGfjzx0yb1PRORWvfligP8siGRh8d1YHDbcKvDEZFKGIbBE+d34oaBLfngjz383+87rQ5JXJCSdKlVb8zZQbHdweMTOlkdikid8vP24IVLurEnJZf3Fuy2OhwRacDW7E/j+elbGdUhgntGtLU6HBGpgmEYPHtRF67qG8Xb83by7oJdVockLkZJutSauMNZfLc6nhsGtqJ1eIDV4YjUuRHtI7ioR3PeX7ibXck5VocjIg1QfFoed36+hhZhfvzf1T01OKtIPWGOYdOdS3o259+zt/PhH3usDklciJJ0qRVOp5MXf91GkK8X94+JtTocEcs8dUFnfL3MudM1kquI1KSsgmJunbyKEoeTj2/qS6i/t9Uhichp8LAZvHZlD87v1owXf9vGO/PV9F1MStKlVszZmsTinSncP6adKg3i1iKCfHji/E6s3JvGF8v3Wx2OiDQQJXYHf/lqLXtTcnn/+t60iQi0OiQROQOeHjbemtiTS3u14LU5O/j3bA0mJ0rSpRbkFpYw6ZctdGwaxI2DWlodjojlruobzYj2Ebw8M469KblWhyMi9ZzT6eSpnzezeGcKL1zSVQPFidRznh42Xr+yB9f0j+bdBbt5fsY2JepuTkm61Lj/m7uDQ5kFvHhpV7w89CsmYhgGr1zeHS8Pg0e+34Bdzd5F5Cy8Mms736yM596RbZnYP8bqcESkBthsBv+6tBs3D27FJ0v38sS0zeom58aUQUmN2pKYyad/7uOa/tH0aakp10TKNA3x5dmLu7BmfzofLdbgMCJyZt5fuJv/LtrNdQNieHR8B6vDEZEaZBgGz1zYmXtHtuXrFQd45PsNFNsdVoclFlCSLjWmsMTOw99tIMzfi7+f29HqcERcziU9WzC+SySvz9nB1sQsq8MRkXrm6xUHeGVWHBf2aM5zF3fFMDSSu0hDYxgGj53bkUfGtefHdQnc8+UaCortVocldUxJutSY/5u7k7jD2bx8WXcNFidyEoZh8OKl3Qj19+K+r9eSU1hidUgiUk98tyqeJ6ZtYlSHCN64qgcemmpNpEG7b3Q7nr+4C/Pikrnx45VkFRRbHZLUISXpUiNW7Uvjgz92M7FfNGM7R1odjojLCg/04a2JvdiXmssTP23SwDAiUqWvVxzgsR82MqxdBO9f30fjvYi4iRsGteLtib1YF5/O1R8sJzm7wOqQpI7ov7yctYy8Ih74dj1RYX48eUFnq8MRcXmD2jbmgbHt+Xl9IlNWxVsdjoi4sC+W7eOfP5l30P93Qx98vTysDklE6tCFPZrz0U392JeSy5X/XUZ8Wp7VIUkdUJIuZ8XhcPLAlPUcyS7kP9f0JtDH0+qQROqFv4yKZUhsY575ZQtbEjOtDkdEXNCnS/fy1M9bGNspkv8qQRdxWyPaR/DVHQPIzC/m8vf/JO6wxrVp6JSky1l5Z8EuFm4/wtMXdqZndKjV4YjUGx42gzev7kWYvzd3fr6GlJxCq0MSERfhdDp5b+Eunp2+lfFdInnvut74eCpBF3FnvWPC+P6uQdgMg6v+u4zV+9KsDklqkZL0esrpdPKvf/2L6Oho/Pz8GD58OOvXr6/Wtj///DPdunXD19eXzp07M2XKlGPe37dvH4ZhnPCYOHFi+Tp2u527HnmKx26+hKR3r+Nv5/dm3LhxrFq1qiZPU6TBmTt3Ltdccw2tWrWiSbAvvVLmkpJTyL1frqWopPJpVqZOncrgwYNp3Lgxvr6+dOjQgRdeeIGioqI6jF5ETtd7773H+eefT+PGjTEMg4ULF55yfYfDyXMztvLqrO1c1KM571zbm5m/TscwDPr27Vs3QYuIpSqrq7eLDGLqPYMID/Th+o9XsCAuuXybyZMnn7T+/t///rfS4zz44IMYhsEjjzxS6+ckp0dJej318ssv8/zzz/P3v/+d6dOnExgYyNixYzl8+PApt1uyZAmXX345o0aNYubMmZx//vlcc801zJkz54R1X3vtNZYtW1b+eOGFF8rf2xafwsfvvknTtl34+ssv+PLLL/Hy8mLo0KGsWbOmxs9XpKGYNWsWGzduZMyYMfj7+xMZ7MurV3Rn5b40nvllS6UDyaWmpjJ69Gg++ugjZs6cya233sqLL77IQw89VMdnICKn4/PPPyctLY3x48dXuW5RiYMHpqzn06X7uGVIK968uif24iIefPBBIiM1KKuIO6iqrh4V5s93dw8itkkgd3y+mmnrEo7Zfv78+cfU3y+77LKTHmfr1q18/PHHBAcH1/o5yekzqhhZWMMOu6CCggIiIyN5+OGHefrppwHIzc2lVatW3HXXXcck08cbP348xcXFzJ8/v3zZeeedR1ZWFkuWLAHMO+mtW7dm+vTpXHDBBSfsIz23iEve+YP0zCymPzyelo0DACgqKqJ9+/aMGjWKTz/99PROatGr5vOIx05vO5F6xuFwYLOZ10fDw8O57777mDRpEq/OiuO9hbt58vxO3D6sTbX29cQTT/Duu++Snp6u+ZLdT3W/cJXjFiv7m9+8eTPdunVjwYIFjBw58oT1cgpLuOfLNSzemcJj53bgnhFtMQyD559/njlz5tC2bVs2b97M6tWr6/4kRGqK6ntVqk5dHSC7oJg7P1/Dsj2pPHl+Jzx2LeLWW28lOzubwMDAKo8zZswYBg8ezBdffMEVV1zBa6+9VivnI5U6ZTmuO+n10J9//klWVhZXXXVV+bKAgAAuvPBCZs6cWel2hYWFLFiw4JjtACZOnMiyZcvIzKx68KrCEjt3fbGGQ9nFfHrXyPIEHcDb25suXbqQmJh4Bmcl4h7KEvTjPTKuAxO6NuWFX7fx8/qEk65zvMaNG6u5u4iLq+xvvqJDmflM/N8ylu5K4dXLu3PvyFgMw+DAgQO8+uqrvPXWW3UQqYhY7XTq6kG+Xnx6S78KdYfq17+nTp1KXFwcjz/+eI3FLjVLSXo9FBcXh4eHB+3atTtmeadOnYiLi6t0u927d1NcXEzHjh1P2M7hcLBjx45jlt9yyy14eHjQrFkzHnroIfLy8nj8h02s3JfGa1f2oE/LRsesX1hYyNq1a2nfvv1ZnqGI+7HZDP7v6p4MaN2IR77fwB87jpx0PbvdTl5eHkuWLOHtt9/mnnvu0V10kXpsfXwGF7+zlL1Hcvnwxr5c1S+6/L2HH36Yq666it69e1sYoYjUldOtq/t6efDutb25Y1hrluxKAaBN27Z4enrSoUMHPvjggxOOkZ+fz8MPP8zLL79MQEDACe+La9B8WfVQeno6gYGBeHgcO9JrWFgYeXl5FBUV4e3tfdLtAEJDQ0/YruL7Pj4+/OUvf2HcuHEEBwezcOFCXnnlFeYu30D28Id4+Jz2XNSj+Qn7f/HFF0lLS+O+++6ridMUcTu+Xh58eFNfrvrvMu7+cg1f3zHwhFkTAgICKCw0R4K/8cYb+fe//21BpCJSE37ZkMij328gIsiHz28bTMemR/uGzp8/nzlz5pxQKReRhqu6dfWKbDaDJ87vTMaOnnyUfT1tuvbirTFtmfXLj9x9993k5eXx4IMPlq//0ksv0axZM66//vraOxE5a0rSXZjT6cRut5e/NgzjhMS8NjRr1ox33nmn/PXIkSM5VOTN/156ghsuuYv7RseesM2vv/7Kiy++yOuvv06HDh1qPUaRhirY14vPb+3P5f/9kxs+WsFnt/Wnd0xY+ft//vkneXl5rFy5kueee4777ruP9957z8KIReR0y+uiEgevzIrj4yV76dcqjP9e34fGgT7l75eUlHD//ffzxBNPaMA4EamWfz94E+PHn8tfv1nHOzu9+fTVdygoKOCFF17gb3/7Gzabjb179/Laa6+xYMECtcJzcWru7sIWLVqEl5dX+WPMmDGAeTUtJyfnmAoBmFfX/P39T3oXvWw74IS+52VX5creP97C7cnMyWsNwNCwrBP+qFetWsXVV1/N3XffzQMPPHB6JykiJ2gS7Mu3dw6iUaA3N3y0glUV5kLt3bs3Q4cO5aGHHuLtt9/m/fffZ/fu3RZGKyKVldcnk5xVwJUfLOPjJXu5aVBLvrx9wDEJOsCHH35IZmYmN998MxkZGWRkZFBUVITdbicjI4Pi4uLaPiURscCZ1tXLjO0cyXd3DaLI7uCy9/4kdsBY0tLS2LdvHwCPP/44EyZMoEOHDuX/WxwOB4WFhWRkZFQ6w4zUPSXpLqxPnz6sWrWq/FHWr6Rjx47Y7XZ27dp1zPpxcXEn9GGpqG3btnh5eZ3Qbz0uLg6bzXbSvuTrDqRzz5draRthjhLp5XnsnYEdO3Zw/vnnM2bMGN5+++0zOk8ROVGLUD+m3DmIyBBfbvx4JUtL+5pVVNZPde/evXUdnohUUFl5fTKP/7iRPUdyeP+63jx7cVd8PE+84759+3YOHjxIZGQkYWFhhIWF8c0337B+/XrCwsLK50wWkYblTOrqx+sWFcLPfxlC24gA/vfHHgAcpbn39u3b+fHHH8v/r4SFhREfH88777xDWFgYCQnVG7hWap+SdBcWFBRE3759yx9lzcgHDx5McHAw33//ffm6eXl5TJ8+nQkTJlS6Px8fH0aNGnXMdgBTpkxh0KBBhISEHLN8V3IOt05eRUSQD+MD9gFmRaTMoUOHGD9+PG3btuWbb76pk6b4Iu6kaYgv3945kJhG/tz0yUq+WxV/zPtLly4FoHXr1laEJyKlKiuvy6TlFvHir1sBaBbix69/HcaEbs0q3d99993HggULjnmMHz+e9u3bs2DBAs4555xaPR8Rscbp1tUr0zzUjyl3DSL8yDpsfsH8Y/Yh4tPy+Oijj0743xIZGclVV13FggULiIiIqI3TkjOgPun1kK+vL48//jjPP/88YWFhdOzYkTfeeAOHw8Ff//rX8vU+//xzbr31Vnbv3k3Lli0BeOqppxg5ciQPPPAAl1xyCb/99hu//fYbs2bNKt9u0qRJHE5JZ2FGIxxevgxrnM7T777FZZddRvfu3QFzZMgJEyaQnp7OO++8w8aNG8u39/HxoVevXnX0aYjUL/v372fVqlUAFBUVsXXrVqZOnUpAQED5Rbb9+/fTtm1bPvnkE2688Ua+u3sQnfoN565lnfltVH8m9m/FsmV/8vrrr3P11VfTtm1bK09JRCrhdDp5e8ps3pq2lOzUwwCMCU1l5YLfSG7Vir59+wInltexsbHExh47/svkyZNJSUk56RzrItJwVKeufnw9AeDyyy+nf//+dO/eHbvdzpQpU4j7cza3PfocK5JymPDWYp6+sDNXjuhzTNdVX19foqOj9b/FxShJr6cef/xxHA4HL730EqmpqfTt25e5c+ceM8CMw+HAbrcf079k6NChTJ06lSeffJL333+f1q1b8/XXXzNu3LjydZrFtOGNj18k98hBbI5ifouJ4dFHH+WJJ54oXycpKYkNGzYAcMEFFxwTW8uWLcv7vojIsRYsWMAtt9xS/vr777/n+++/P+bvpmwQKofDAUCInxe3XDyGD7/4ls+Xfs03np50bB/LSy+9xN13323FaYhIFfal5PLs9C18/8Zr5G6eV778+eeeBeCmm25i8uTJwMnLaxFxT9Wpqx9fTwDo0KEDn3zyCfHx8TidTjp37sznn3/ODTfcwMH0PB76bgOPTd3Ij2sP8sIl3YhtEmjF6Uk1GVUUCCot3ExyVgETP1xOUmYBn93an76tGlW9UU1Y9Kr5POKxujmeSD3kdDr5Yvl+np+xlfBAH96+phf96upvVFxNdYflVTlex/KKSnhvwW7+98cevD1tPDC2HTcNboWXh3oYiqi+Zx2Hw8m3q+J5ZVYceUUl3Dy4FfeOjCUs4OQDTkutO2U5riRdyiVnF3DN/5ZzqDRBr9PKv/5pi1TbpoOZ3PfNWg6m5/O3Me24d2RbPJUAuBsl6S7G4XAyfWMir8yMIzGzgMt6teDxCR1pEuxrdWgirkP1Pcul5BTyysw4flh7kABvT+4Y3oabBrUixN/L6tDcjZJ0qdqB1Dxu/GQFydmFTL6lP/1b1/HdOf3TFjkt2QXFPDltMz+vT6RbixD+fWV3OjYNtjosqTtK0l2E0+lk4Y4jvDprO9sOZdG5WTDPXtxFrVxETkb1PZexMymb1+ZsZ/aWJAK8Pbi6Xwy3Dm1FVJi/1aG5CyXpcmpbEjO56ZNVlDgcfHxTP/q0PPUcjLVC/7RFzshvmw7x1LTNZBUUc//odtw9sq2a1boHJekuYO2BdF6ZGceKvWnENPLn4XHtubB7c2y26n49Im5G9T2XszUxi48W7+GXDYk4gXO7NOW6gTEMatP4mAHmpMYpSZfKzdp8iIe/20CInxef39af2CZB1gSif9oiZywtt4hnftnC9A2JdGkezKtXdKdL8+pN0yL1lpJ0C+1Iyub10jtQ4YHe3D+mHRP7xeDtqQtkIqek+p7LSszIZ/Kf+5iyKp7M/GLaRARw3YCWXNE7Sk3ha4eSdDmR3eHkjbnbeXfBbnpGh/Lf6/vQNMTCfnP6py1y1mZtPsST0zaTnlfMzYNb8eA57Qn00SQeDZSSdAtsSczknfm7mLn5MIE+ntw1vA23Dm1NgP7ORKpH9T2XV1Bs59eNh/hyxX7WHcjAx9PGBd2bc1XfKPq3bqS76zVHSboca8+RHB75fgNrD2RwTf9oJl3UBR9PD2uD0j9tkRqRmVfMq7Pj+HrlASKDfHnmws6c27WpCtWGR0l6HdoQn8E7C3Yxd2sSQT6e3DKkFbcMaa1RkUVOl+p79cqWxEy+WnGAX9YnklNYQqvG/lzZN5rLeregWYif1eHVd0rSxVRYYufTpft48/cd+Hh68NzFXbi4ZwurwzLpn7ZIjVp7IJ0nftrMtkNZjOoQwZMXdKZthOZEbUCUpNeyYruDWZsPM/nPfazZn06Inxe3DW3NTYNbEeKnpp8iZ0T1vXopr6iEWZsP893qeJbvScNmwLB2EVzVN5qxnZtYf7OvflKS7u7sDie/bTrEq7PjiE/LZ2ynJrx4aTciXWlaGP3TFqlxJXYHk//cx//N3UFBiYNr+8fwt7HtCA/0sTo0OXtK0mtJQkY+P645yJcr9pOUVUirxv7cOKgVV/aNIshXybnIWVF9r97bn5rL1DUH+WHNQRIzCwjy8WRs50jO69aMYe3C8fVSwl5Nrpekv7nmzdrYrRynqNhg24Fg1u0IJSPHm/CQQob1SKFlZJ7VoZ2g/46FAKxsP9LSOMS1PNDnAatDaBCOZBfy1rwdfLMyHj8vD64f2JKbB7eydhwKOVtK0mtQclYBMzcf5pcNiazZnw7AsHbh3DqkNSPaR2i09lqmeqH7UH2v4XA4IT7Jn+3xgexOCKSw2AMvTwetm+XSMjKPmMg8nhp2n9VhujLXS9Lvm6cvrLY4HAYpaY05eKgFh5MjcTg8CA3JoG3LPTRtkoSrdks998BGAGbFdLc4EnEl74x5x+oQGpRdyTn839wdzNx8CA+bwUU9WnDdwBh6RYeqz3r9oyT9LBSVONiUkMHC7UeYH5fMlsQsADpEBnFRz+Zc2L05MY01V3BdUb3Qfai+1zA5HAap6Y05lBRJ0pFICovMFnttIwIY0KYxvaJD6d0yjNaNA3TR86hTfhAajrSeczohNy+AtIwwjqSGcyQ1nJISL7y8iohufpCoZomEhmS4bHIuInUntkkg717XmwOpeXyydC9TVsXzw9qDtGzsz8U9W3Bul6Z0ahakhF0anCPZhWw8mMHq/ems2ZfOhoMZFJY4sBnQp2UYj47vwNhOkXRoatE0pCIi9ZjN5iSicQoRjVNwOreQnRPEkdTGNDGGM31DIl+vOABAiJ8XPaNDyx/do0JorC54J6U76fWI0wkFhb5kZQeRnRNEemYI6RlhFBWbv9w+3gU0iThCZHgyTcKPYLO53g2UlZ+tZNUXq05Y/uRwc4TcF/4oquuQRGrFM888w6RJk6wO45SyC4qZtfkw09Yn8OfuVJxOiAjyYVhsOAPbNKZnTChtIwLx0FVvV6Q76SdhdzjZl5rL1sQsth7KKn8+kl0IgKfNoEuLEPq2DKNvyzAGtw1vUPP/Tpo0iWeffdbqMEQqpfreUf1u6Ef/m/pbHUatemfMOzgcTvak5LD2QAbrDmSw7kA6O5KycZSWTtGN/OgRdTRx79I8BD9vt+jXrjvp9YnDYVBQ6Etenj95+X7k5R99zs0LoLjkaGXC3y+XJhFHaBSSTqOwdAL8c3XHXESqLcjXiyv7RnNl32iSswpYtOMIi3emsHDHEX5clwBAoI8n3VqE0Ll5MO2aBBLbJJB2TYIaVGIj9VNmfjG7knOIO3w0GY87lE1+sR0wE/J2kUEMbxdB5+bBdGkeTI+oUHep/ImIuASbzSC2SRCxTYK4qm80ALmFJWxKyGRDfAYbDprJ+4yNhwDwsBl0bBpEr5hQeseE0SsmjFaN/d2ulZ+SdAs4nZCX70duXkDpw7/85/wCX5xOW/m6huHAz7cAf788mjc9RFBgNsFBWQQF5ODlVWLhWYhIQ9Ik2Lc8YXc4nOxNzWVDfAbrSx9frdhPQbGjfP3wQB+iwvxoEepH81BfmoX40TzUj2YhvkQE+RAe6IO3p+0URxSpWn6RncTMfBLS89lzJIddR3LYlZzDruRcUnIKy9cL8vWkc7NgJvaPpnOzYDo3Dya2SaCmBRIRcUEBPp4MbNOYgW0aly9Lzi5gQ3wm6+PTWR+fwU9rE/hyudlMvlGAN72iQ8sT9+7RoQT6NOw0Vs3da4nTCYWFPuTmBZBTIQnPzQsgL88fR4VE3NOzmAD/XAL88/D3y8PfL7/82denwCWbrdc0DSQiJ6OB41yHw+EkISOfncnZ7EzKYfeRHBIzCkjMyCcxM/+YBL5MmL8XEUE+NAnyLX32IaLCw3ztS7CvZ4O4Ql5id5CWV0RGXjGFxQ6K7HbCA31o2TigNg5nWXP3fSm5pOYW4XQ6cWKWdw6nE6cTnJQ+l/7scFJhvYrvmduU2J1k5heTkV9EZl4xGXnFpOUVcSgzn8SMAtJyj20SG+TrSWyTQGIjAmlb+tyhaRBRYX4N4nfIXblDvVBMqu+5l7Opx9kdTnYmZ7N2v9lEfu2BdHYfyQXAZkD7yCB6xYTRPSqEVo0DaBXuT2SQb30amM71mrvHhsZacdgzVlxikFvgicMBdoeB3WHgcBgU220UFNooKPKgoMiDnHwPsnK9zEeeJ3bH0UTcw+YgNLCYyNBiwqIyCAsqJjSwiNCgYvx97Mc1U/ctfbiPRskHgfr3uyHiLmw2g+hG/kQ38md0x8hj3nM6naTnFZOYkU9SVgFHsgtJzi4sfTZfr9qXS3J2IUUlJybz3p42IgJ9aBLsU+H5aGIf6u+Fn7cHfl4e+Hl74O/lia+3DW8PW60lZk6nk7wiO+l5RaTnFpOaW0h6XhGpOUWk5xWRlmv+nJZb+nNuEZn5xSfs56ZBLXn24q61EqNV3pi7g182JNb4fr09bYT5exHm703TEF96RIXSPLSstYYfrcL9iQj0UTLeAKnsdx+q70l1mc3eg+nYNJhrB8QAkJlXzLr4dNYdyGDtgXRmbEzkm5UHyrfx9bIR08if8EAfwgK8aeTvTViANwHeHnh72vDxNJ+9PW142oxjsuRji5ajL3rHhNIkuO7zMkvupNc3szYf4u4v11a5Xqi/F9Fh/kQ38iMqzJ+oMD9aNQ6gdXgAzUP9NPjSqaz/xnzueY21cYhIrXE6nWQXlpCcdWwCf6Q8oS/9OafwhDuoJ2MzwN/bE18vD/y8bfh6euDpYcPbw8DTwyyAywpiLw/zoqnD6aTE4cRe+ihxOCm2O8gvspNbVEJ+kZ28Ijv5xXYqKx49bAaNArxpHOBNo9KH+bMPjQK9CfP3wtfTAy9PG1FhfrSNCKzJj7GMZXfStyZmcSSnEAOzUmMzSis6FX42DAObUVbpMY5Zr+xnAE8Pg1A/b0L9vfD1UtN0kQZP9T2pQQ6Hk8TMfPal5LEvNZd9KbkcSMszL6rnFpGeW0RGfnGl5Xl1fHhjX87pHFn1iqfP9eZJr28SMvJZsScVLw8bXh5G6bMNXy8Pwvy9CPX3JsTPS/0vRURqSFGJg9TcQpKzCsnMLya/2E5+afJ8/HNekZ2CYvNRbHdS4nBQbHdQbDcT8JLSZzATbE+bga302aM0gffz8sDf2wM/b0/8vT1KH540CvAyk++yZ39vgv1conm+RncXERGpgt3hpLDEXtoNzUFhsYPCEjv2CjlwxXT4+NQ4qpEfwb61MliuknQREZEGRkm6iIhI/XXKcly3fkVERERERERchJJ0ERERERERERehJF1ERERERETERShJFxEREREREXERStJFREREREREXISSdBEREREREREXoSRdRERERERExEUoSRcRERERERFxEUrSRURERERERFyEknQRERERERERF6EkXURERERERMRFGE6ns9I3n3322VlAeC0ctzmQWAv7rY/0WRylz8Kkz+EofRZH6bMw6XMwpTzzzDPnVrVSLZbjdcEdv2t3PGfQebsbdzxvdzxn0HmfyqnLcafTWeePSZMmOa04ris+9Fnos9DnoM9Cn4U+Bz30XbvrOeu8rY9D561z1nm73nmrubuIiIiIiIiIi7AqSX/WouO6In0WR+mzMOlzOEqfxVH6LEz6HNyHO37X7njOoPN2N+543u54zqDzPmOn7JMuIiIiIiIiInVHzd1FREREREREXISSdBEREREREREXoSRdRERERERExEXUSZJuGMadhmEsMAwjwzAMp2EYraq53eWGYWw1DKOw9PnSWg611hmG4WMYxn8Mw0gxDCPXMIxfDMOIqmKbSaWfW8XH4bqKuaYYhnGvYRh7DcMoMAxjjWEYw6pYf0TpegWGYewxDOPuuoq1Np3O52AYxsiTfPdOwzA61mXMtcEwjOGlv/8Jped0czW26WYYxiLDMPJLt3vaMAyjDsKtNaf7ORiG0aqS34kq58x2ZYZh/MMwjFWGYWQZhnHEMIzphmF0rcZ2De53wt0YhtGotFyMK/0e4w3DeN8wjMbV2LZe1xPOpH5kGMbNlfwP8K2DkGuEO9YL3aX+5651PXer27lrHa4u62x1dSfdH5gDTKruBoZhDAKmAF8BPUufvzcMY0AtxFeX3gQuB64BhgHBwAzDMDyq2G470KzCo1stxljjDMO4GngL+BfQC/gTmGkYRkwl67cGfitdrxfwEvAfwzAur5uIa8fpfg4VdOHY739nbcZZRwKBzcDfgPyqVjYMIxiYCyQB/Uq3exR4qBZjrAun9TlUcC7H/k7Mr/nQ6tRI4D1gMDAaKAF+NwyjUWUbNODfCXfTHGgBPIZZtl0PDAe+OdVGDaSecNr1o1J5HPv338zpdBbUbGi1yh3rhW/SwOt/7lrXc9O6nbvW4equzlaXE7sDfQEn0Koa604B5h637HfgG6snqD+L8w8BioDrKiyLBhzA+FNsNwnYbHX8Z3nuK4APj1u2E3ipkvVfAXYet+wjYJnV51LHn8PI0r+ZcKtjr+XPJQe4uYp17gGyAL8Ky54EEiidqaK+P6r5ObQq/Z3oa3W8tfxZBAJ24EJ3/p1w1wdwXmnZGHyKdRpMPeE060c3AzlWx2zBedfb79td6n/uWtdz97qdu9bharvO5sp90gdhXmWtaDbmXZb6qg/gRYXzcjqd8cA2qj6vNoZhJJY2pfnWMIw2tRhnjTIMwxvz3I//PudQ+XlX9v33NQzDq2YjrBtn+DmUWW0YxiHDMOYZhjGqVgJ0fYOAxU6ns+KVy9mYd+FaWRKRtX40DCPZMIylhmFcYXUwtSAIs7VX+inW0e9EwxUMFGLeMa5MQ6wnVJefYRj7DcM4aBjGDMMwelkdUB2oz993g6//uWtdT3W7anP38vq062yunKQ3xWwSUVFS6fL6qinmnaGU45ZXdV4rMK+cnwvcUbrun0Y1+uu5iHDAg9P7Piv7/j1L91cfncnncAjz6uPlwGWYzd7mVdXHq4Gq7Hei7D13kQM8AlyFebdxHjDFMIzrLY2q5r0FrAeWnWId/U40QIZhhALPY96ZKjnFqg2xnlAd24FbgYsxm04XAEsNw2hnaVS1rz5/3+5Q/3PXup7qdtXjruX1GdfZPM/0iIZhvAA8UcVqo5xO58IzPUZ9Ud3P4kz373Q6Zx53vOXAHuAm4I0z3a+4PqfTuR3zn3eZZYY5wM6jwGJLghJLOZ3OFOD1CotWG4YRjtmf90troqpZhmG8AQwFhjqdTrvV8ciZOZN6gmEYgcB0zCaQj9VedLWntutHTqfz/9u7f1A5ijiA499B0BQRQYSXQg2x0ec/XmW0MVFyKMFCVIiIoIKFiErEJjaaTkQhF0TSyZFCApJK0SokWHgEXgQLJSDoU0TFf0WIhZfIWMw83uZ8f+72/uze7fcDB9nb2Xv7m92b/c3kdrZLYfAqhPAFaUDrZeCVMp85Dk3MC83/VIa5XXOMkrOV7qSTJsDYKiH8cYTP/xVY6HtvIb9fN20Gq4t7SaNtNwC/F9YtMMSXMsZ4MYTwNTAro+Z/kEaQhzmeGx3/y/x/JHpWlKmH9ZwFnhzXTs2Qjc6J1XVNdhZ4ruqdGIcQwhHS+f1AjPG7LYp7TtRbmyHyhNxB/zQvPhK3ngitrnlCm8nmR1eIMf4bQlim+pygTfPywjbmf6uamuuZ2w3G6/WagXK20p30PDIwyS9QF2gB7xTea5FmTKyVQesihHAOuESK48P83o3AIkPEFdJjVm4DTpfZ32mLMfZy7C3go8KqFnByg826QP+jVVrAcozx0vj3cvJK1sN6lkg/lWqaLvB2CGFbIXlvAT8DK5XtVT0sMQfnRAjhKHCA1EE/P8AmnhM1NkyeEEK4FvgMCMDDMcaLA2xWyzxhCvnRFfIjjO4GvprW31xPE/NC8781Tc31zO0G5vV6zRKDHOspzX63I+/QU6QZ7vbn5esLZU5RmAWRNNnCZeAQqUF6ndTA7a56Nr8R6+IY8BOwj/SYhtOkn6ldVShzHnipsPwusAfYBewGPiHNkLiz6niGiPsAaWbT50kXpaOk+zR25vXHgeOF8ruAv0mj1It5ux7weNWxTLkeDgKPkkbN7yA9niQCj1UdyxjqYntuB5ZIk0O9kf99c17/FnCqUP460mjrCeBO0n1cF4DXqo5lyvXwTG5LF4FbSfc69YBXq45lxHp4Px/PB0nXjNXX9kKZRpwTTXuRJgnsAqv/Q1g8/lcXys1dnkC5/OhN4CHgllz2gxz3PVXHM+G4Z/p404D8j4bmeiXiPsiM53Y0NIcrEXfpnG1aAR3OJ1//69lCmRWg07fdE7nB6pFmwJyZk3eTurgGeA/4Mx/cj4Gb+spE4HBh+QRppKlHuk/vJHB71bGUiP3FfJz/Ac4B9xfWnQHO9JXfA3yZy38PvFB1DNOuB9I9K9+SnsX4F+lncfurjmFM9bB3g3ahk9d3gJW+be4CPidNlPQLKVmdyUd3lK2H3OB/Q0psLgDLwNNVxzGGelivDvrbwkacE017bfIdiMDeQrkV5ixPoER+BBwBfsjXkN9IMyTfV3Usk4571o83Dcn/aGiuN0zczEFut0m73cnrO8zh9XrYuBkhZwv5AyRJkiRJUsXq/Ag2SZIkSZIaxU66JEmSJEk1YSddkiRJkqSasJMuSZIkSVJN2EmXJEmSJKkm7KRLkiRJklQTdtIlSZIkSaoJO+mSJEmSJNWEnXRJkiRJkmriP2aZoLohLgxGAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 993.6x331.2 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "az.plot_posterior(results, var_names=['treatment','treatment:pucca'], ref_val=0, rope=[-1, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "counts = ddb[ddb.seekhelpfever.notna() & ddb.dwelling.notna()].groupby('stratumid').count().userid.sort_values().reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 909,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import statsmodels.formula.api as smf\n",
    "import statsmodels.api as sm\n",
    "import numpy as np\n",
    "\n",
    "\n",
    "def get_percentages(base, treatment):\n",
    "    return expit(base), expit(base + treatment), 1 - ((1 - expit(base + treatment))) / (1 - expit(base))\n",
    "\n",
    "def reg(form, dat, weights_var='waves_answered'):\n",
    "    try:\n",
    "        model = smf.glm(form, dat, family=sm.families.Binomial(), var_weights=dat[weights_var])\n",
    "    except KeyError:\n",
    "        return smf.logit(form, dat) \\\n",
    "                   .fit(cov_type='cluster', cov_kwds={'groups': dat['stratumid']}) \\\n",
    "                   .summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 208,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb = ddb.merge(dd[['stratumid', 'pre_seekhelpfever']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 217,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.201553\n",
      "         Iterations 7\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>seekhelpfever</td>  <th>  No. Observations:  </th>  <td>  1536</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  1534</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Fri, 09 Apr 2021</td> <th>  Pseudo R-squ.:     </th> <td>0.005712</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>08:42:16</td>     <th>  Log-Likelihood:    </th> <td> -309.59</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -311.36</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th>  <td>0.05930</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    2.7229</td> <td>    0.172</td> <td>   15.860</td> <td> 0.000</td> <td>    2.386</td> <td>    3.059</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.4426</td> <td>    0.284</td> <td>    1.558</td> <td> 0.119</td> <td>   -0.114</td> <td>    0.999</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   No. Observations:                 1536\n",
       "Model:                          Logit   Df Residuals:                     1534\n",
       "Method:                           MLE   Df Model:                            1\n",
       "Date:                Fri, 09 Apr 2021   Pseudo R-squ.:                0.005712\n",
       "Time:                        08:42:16   Log-Likelihood:                -309.59\n",
       "converged:                       True   LL-Null:                       -311.36\n",
       "Covariance Type:              cluster   LLR p-value:                   0.05930\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      2.7229      0.172     15.860      0.000       2.386       3.059\n",
       "treatment      0.4426      0.284      1.558      0.119      -0.114       0.999\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 217,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('seekhelpfever ~ treatment', ddb[ddb.seekhelpfever.notna() & ddb.dwelling.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "smf.ols('bednet_share ~ treatment', data=mat[mat.hasmosquitonet == 'Yes'], var_weights=mat[mat.hasmosquitonet == 'Yes']['waves_answered']).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('bednet_share ~ treatment*pregnantwoman', mat[mat.hasmosquitonet == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('sleepundernet ~ treatment*pregnantwoman', mat[mat.hasmosquitonet == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "matb = ddb\n",
    "\n",
    "matb['seekhelp_if_fever'] = matb['seekhelpfever']\n",
    "matb['fever2weeks'] = (matb['fever4months'] == 1).astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "f = ddb[ddb.fever4months.notna()]['stratumid'].astype('category').cat.codes.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb.groupby('treatment').admnm.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('knowmnm ~ treatment*pucca', ddb[ddb.knowmnm.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def power(n, p):\n",
    "    return n**p"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from mnm.prep import cast_number\n",
    "\n",
    "ddb['age'] = ddb.age.map(lambda a: cast_number(0, 100, a))\n",
    "ddb['young'] = ddb.age < 30"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb['female'] = (ddb.gender.notna() & (ddb.gender == 'Female') )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb[ddb.admalaria.notna() & ddb.female & ~ddb.young].groupby('treatment').admalaria.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb[ddb.admalaria.notna() & ddb.female & ~ddb.young].groupby('treatment').admnm.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('admalaria ~ treatment', ddb[ddb.admalaria.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('admalaria ~ treatment*female*young', ddb[ddb.admalaria.notna() & ddb.gender.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from mnm.prep import cast_number\n",
    "\n",
    "ddb[ddb.admalaria.notna()].age.map(lambda a: cast_number(0, 100, a)).plot.hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb[ddb.admalaria.notna()].gender.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 300,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.500436\n",
      "         Iterations 6\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>        <td>knowmnm</td>     <th>  No. Observations:  </th>  <td>  7939</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  7937</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     1</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Fri, 09 Apr 2021</td> <th>  Pseudo R-squ.:     </th>  <td>0.01952</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>09:05:55</td>     <th>  Log-Likelihood:    </th> <td> -3973.0</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -4052.1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th> <td>2.815e-36</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>   -1.0556</td> <td>    0.514</td> <td>   -2.054</td> <td> 0.040</td> <td>   -2.063</td> <td>   -0.049</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>   -0.7227</td> <td>    0.516</td> <td>   -1.401</td> <td> 0.161</td> <td>   -1.734</td> <td>    0.288</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:                knowmnm   No. Observations:                 7939\n",
       "Model:                          Logit   Df Residuals:                     7937\n",
       "Method:                           MLE   Df Model:                            1\n",
       "Date:                Fri, 09 Apr 2021   Pseudo R-squ.:                 0.01952\n",
       "Time:                        09:05:55   Log-Likelihood:                -3973.0\n",
       "converged:                       True   LL-Null:                       -4052.1\n",
       "Covariance Type:              cluster   LLR p-value:                 2.815e-36\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept     -1.0556      0.514     -2.054      0.040      -2.063      -0.049\n",
       "treatment     -0.7227      0.516     -1.401      0.161      -1.734       0.288\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 300,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('knowmnm ~ treatment', ddb[ddb.knowmnm.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 301,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.408594\n",
      "         Iterations 6\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>         <td>admnm</td>      <th>  No. Observations:  </th>  <td>  7797</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  7795</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     1</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Fri, 09 Apr 2021</td> <th>  Pseudo R-squ.:     </th>  <td>0.04016</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>09:06:01</td>     <th>  Log-Likelihood:    </th> <td> -3185.8</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -3319.1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th> <td>6.294e-60</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>   -1.3271</td> <td>    0.640</td> <td>   -2.073</td> <td> 0.038</td> <td>   -2.582</td> <td>   -0.072</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>   -1.1168</td> <td>    0.644</td> <td>   -1.735</td> <td> 0.083</td> <td>   -2.378</td> <td>    0.145</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:                  admnm   No. Observations:                 7797\n",
       "Model:                          Logit   Df Residuals:                     7795\n",
       "Method:                           MLE   Df Model:                            1\n",
       "Date:                Fri, 09 Apr 2021   Pseudo R-squ.:                 0.04016\n",
       "Time:                        09:06:01   Log-Likelihood:                -3185.8\n",
       "converged:                       True   LL-Null:                       -3319.1\n",
       "Covariance Type:              cluster   LLR p-value:                 6.294e-60\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept     -1.3271      0.640     -2.073      0.038      -2.582      -0.072\n",
       "treatment     -1.1168      0.644     -1.735      0.083      -2.378       0.145\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 301,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('admnm ~ treatment', ddb[ddb.admnm.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.10449954614805741, 0.08705782423611545, -0.019477066523994946)"
      ]
     },
     "execution_count": 212,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(-2.1482, -0.2019)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('admalaria ~ treatment', ddb[ddb.admalaria.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.17265877040255395, 0.13918208036586574, -0.04046297807855748)"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(-1.5669, -0.2552)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('admalaria ~ treatment*pucca', ddb[ddb.admalaria.notna() & ddb.seekhelpfever.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.22422717743519824, 0.16955279551333116, -0.07047731027893822)"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(-1.706 + 0.4648, -0.0661 - 0.2815)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.15368325421926876, 0.14528136778740763, -0.009927590909370787)"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(-1.706, -0.0661)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([nan, 'Semi-pucca', 'Kutcha (made of mud, tin, straw)',\n",
       "       'Pucca (have cement/brick wall and floor'], dtype=object)"
      ]
     },
     "execution_count": 154,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb.dwelling.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 632,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb['kutcha'] = ddb.dwelling == 'Kutcha (made of mud, tin, straw)'\n",
    "ddb['pucca'] = ddb.dwelling == 'Pucca (have cement/brick wall and floor'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('fever ~ treatment', ddb[ddb.malaria4months.notna() & ddb.fever4months.notna() & ddb.seekhelpfever.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('testmalaria ~ treatment*pucca', ddb[ddb.dwelling.notna() & ddb.testmalaria.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 607,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5598006644518272"
      ]
     },
     "execution_count": 607,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb.testmalaria.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.462360\n",
      "         Iterations 5\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>fever4months</td>   <th>  No. Observations:  </th>  <td>  4920</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  4916</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     3</td>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Wed, 24 Feb 2021</td> <th>  Pseudo R-squ.:     </th> <td>0.0002249</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>16:25:52</td>     <th>  Log-Likelihood:    </th> <td> -2274.8</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -2275.3</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th>  <td>0.7956</td>  \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "             <td></td>                <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th>               <td>   -1.5630</td> <td>    0.090</td> <td>  -17.408</td> <td> 0.000</td> <td>   -1.739</td> <td>   -1.387</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>pucca[T.True]</th>           <td>    0.0369</td> <td>    0.116</td> <td>    0.317</td> <td> 0.751</td> <td>   -0.191</td> <td>    0.265</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th>               <td>   -0.0474</td> <td>    0.108</td> <td>   -0.439</td> <td> 0.661</td> <td>   -0.259</td> <td>    0.164</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment:pucca[T.True]</th> <td>    0.0628</td> <td>    0.164</td> <td>    0.383</td> <td> 0.701</td> <td>   -0.258</td> <td>    0.384</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:           fever4months   No. Observations:                 4920\n",
       "Model:                          Logit   Df Residuals:                     4916\n",
       "Method:                           MLE   Df Model:                            3\n",
       "Date:                Wed, 24 Feb 2021   Pseudo R-squ.:               0.0002249\n",
       "Time:                        16:25:52   Log-Likelihood:                -2274.8\n",
       "converged:                       True   LL-Null:                       -2275.3\n",
       "Covariance Type:              cluster   LLR p-value:                    0.7956\n",
       "===========================================================================================\n",
       "                              coef    std err          z      P>|z|      [0.025      0.975]\n",
       "-------------------------------------------------------------------------------------------\n",
       "Intercept                  -1.5630      0.090    -17.408      0.000      -1.739      -1.387\n",
       "pucca[T.True]               0.0369      0.116      0.317      0.751      -0.191       0.265\n",
       "treatment                  -0.0474      0.108     -0.439      0.661      -0.259       0.164\n",
       "treatment:pucca[T.True]     0.0628      0.164      0.383      0.701      -0.258       0.384\n",
       "===========================================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('fever4months ~ treatment*pucca', ddb[ddb.dwelling.notna() & ddb.fever4months.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('fever4months ~ stratumid', ddb[ddb.dwelling.notna() & ddb.fever4months.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.17321658792823388, 0.16653308624261706, -0.00808373945102403)"
      ]
     },
     "execution_count": 176,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(-1.5630, -0.0474)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  seekhelpfever\n",
       "0          0.0              501\n",
       "1          0.0              464\n",
       "Name: seekhelpfever, dtype: int64"
      ]
     },
     "execution_count": 180,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[ddb.seekhelpfever.notna()].groupby('treatment').seekhelpfever.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "smf.logit('seekhelpfever ~ treatment', ddb[ddb.seekhelpfever.notna()]).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 639,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.203132\n",
      "         Iterations 7\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>seekhelpfever</td>  <th>  No. Observations:  </th>  <td>  1483</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  1481</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Mon, 12 Apr 2021</td> <th>  Pseudo R-squ.:     </th> <td>0.004916</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>14:17:52</td>     <th>  Log-Likelihood:    </th> <td> -301.24</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -302.73</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th>  <td>0.08449</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    2.7249</td> <td>    0.181</td> <td>   15.050</td> <td> 0.000</td> <td>    2.370</td> <td>    3.080</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.4091</td> <td>    0.291</td> <td>    1.408</td> <td> 0.159</td> <td>   -0.161</td> <td>    0.979</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   No. Observations:                 1483\n",
       "Model:                          Logit   Df Residuals:                     1481\n",
       "Method:                           MLE   Df Model:                            1\n",
       "Date:                Mon, 12 Apr 2021   Pseudo R-squ.:                0.004916\n",
       "Time:                        14:17:52   Log-Likelihood:                -301.24\n",
       "converged:                       True   LL-Null:                       -302.73\n",
       "Covariance Type:              cluster   LLR p-value:                   0.08449\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      2.7249      0.181     15.050      0.000       2.370       3.080\n",
       "treatment      0.4091      0.291      1.408      0.159      -0.161       0.979\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 639,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('seekhelpfever ~ treatment', ddb[ddb.seekhelpfever.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8903772184388211, 0.9271828978731256, 0.3357484540178699)"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(2.0946, 0.4496)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8664582774667876, 0.9228346818824007, 0.4221632261901672)"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(1.87, 0.6115)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 476,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>OLS Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>      <td>seekhelpfever</td>  <th>  R-squared:         </th> <td>   0.003</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                   <td>OLS</td>       <th>  Adj. R-squared:    </th> <td>   0.002</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>             <td>Least Squares</td>  <th>  F-statistic:       </th> <td>   3.868</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>             <td>Thu, 18 Mar 2021</td> <th>  Prob (F-statistic):</th>  <td>0.0495</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                 <td>16:07:52</td>     <th>  Log-Likelihood:    </th> <td> -33.676</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Observations:</th>      <td>  1186</td>      <th>  AIC:               </th> <td>   71.35</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Residuals:</th>          <td>  1184</td>      <th>  BIC:               </th> <td>   81.51</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Df Model:</th>              <td>     1</td>      <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>nonrobust</td>    <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>t</th>      <th>P>|t|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    0.9195</td> <td>    0.010</td> <td>   91.079</td> <td> 0.000</td> <td>    0.900</td> <td>    0.939</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.0285</td> <td>    0.014</td> <td>    1.967</td> <td> 0.049</td> <td>  6.8e-05</td> <td>    0.057</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "  <th>Omnibus:</th>       <td>842.204</td> <th>  Durbin-Watson:     </th> <td>   1.958</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Prob(Omnibus):</th> <td> 0.000</td>  <th>  Jarque-Bera (JB):  </th> <td>7324.192</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Skew:</th>          <td>-3.459</td>  <th>  Prob(JB):          </th> <td>    0.00</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Kurtosis:</th>      <td>13.018</td>  <th>  Cond. No.          </th> <td>    2.59</td>\n",
       "</tr>\n",
       "</table><br/><br/>Notes:<br/>[1] Standard Errors assume that the covariance matrix of the errors is correctly specified."
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                            OLS Regression Results                            \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   R-squared:                       0.003\n",
       "Model:                            OLS   Adj. R-squared:                  0.002\n",
       "Method:                 Least Squares   F-statistic:                     3.868\n",
       "Date:                Thu, 18 Mar 2021   Prob (F-statistic):             0.0495\n",
       "Time:                        16:07:52   Log-Likelihood:                -33.676\n",
       "No. Observations:                1186   AIC:                             71.35\n",
       "Df Residuals:                    1184   BIC:                             81.51\n",
       "Df Model:                           1                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      0.9195      0.010     91.079      0.000       0.900       0.939\n",
       "treatment      0.0285      0.014      1.967      0.049     6.8e-05       0.057\n",
       "==============================================================================\n",
       "Omnibus:                      842.204   Durbin-Watson:                   1.958\n",
       "Prob(Omnibus):                  0.000   Jarque-Bera (JB):             7324.192\n",
       "Skew:                          -3.459   Prob(JB):                         0.00\n",
       "Kurtosis:                      13.018   Cond. No.                         2.59\n",
       "==============================================================================\n",
       "\n",
       "Notes:\n",
       "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n",
       "\"\"\""
      ]
     },
     "execution_count": 476,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smf.ols('seekhelpfever ~ treatment', ddb[(ddb.week < 20) & ddb.dwelling.notna() & ddb.seekhelpfever.notna()]).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 459,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.243136\n",
      "         Iterations 7\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>seekhelpfever</td>  <th>  No. Observations:  </th>  <td>  1186</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  1184</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Thu, 18 Mar 2021</td> <th>  Pseudo R-squ.:     </th> <td>0.006724</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>16:02:33</td>     <th>  Log-Likelihood:    </th> <td> -288.36</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -290.31</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th>  <td>0.04817</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    2.4361</td> <td>    0.168</td> <td>   14.526</td> <td> 0.000</td> <td>    2.107</td> <td>    2.765</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.4671</td> <td>    0.288</td> <td>    1.622</td> <td> 0.105</td> <td>   -0.097</td> <td>    1.032</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   No. Observations:                 1186\n",
       "Model:                          Logit   Df Residuals:                     1184\n",
       "Method:                           MLE   Df Model:                            1\n",
       "Date:                Thu, 18 Mar 2021   Pseudo R-squ.:                0.006724\n",
       "Time:                        16:02:33   Log-Likelihood:                -288.36\n",
       "converged:                       True   LL-Null:                       -290.31\n",
       "Covariance Type:              cluster   LLR p-value:                   0.04817\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      2.4361      0.168     14.526      0.000       2.107       2.765\n",
       "treatment      0.4671      0.288      1.622      0.105      -0.097       1.032\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 459,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('seekhelpfever ~ treatment', ddb[(ddb.week < 20) & ddb.dwelling.notna() & ddb.seekhelpfever.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 460,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8175744761936437"
      ]
     },
     "execution_count": 460,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(1.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 458,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9190865327845347"
      ]
     },
     "execution_count": 458,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(2.43)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 471,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.296137339055794"
      ]
     },
     "execution_count": 471,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[ddb.seekhelpfever.notna() & ddb.dwelling.isna()].seekhelpfever.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 467,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9333895446880269"
      ]
     },
     "execution_count": 467,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ddb[ddb.seekhelpfever.notna() & ddb.dwelling.notna()].seekhelpfever.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 473,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.457360\n",
      "         Iterations 6\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>seekhelpfever</td>  <th>  No. Observations:  </th>  <td>  1418</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  1416</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Thu, 18 Mar 2021</td> <th>  Pseudo R-squ.:     </th> <td>0.001504</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>16:07:11</td>     <th>  Log-Likelihood:    </th> <td> -648.54</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -649.51</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th>  <td>0.1622</td> \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    1.4833</td> <td>    0.094</td> <td>   15.726</td> <td> 0.000</td> <td>    1.298</td> <td>    1.668</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.1976</td> <td>    0.157</td> <td>    1.259</td> <td> 0.208</td> <td>   -0.110</td> <td>    0.505</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   No. Observations:                 1418\n",
       "Model:                          Logit   Df Residuals:                     1416\n",
       "Method:                           MLE   Df Model:                            1\n",
       "Date:                Thu, 18 Mar 2021   Pseudo R-squ.:                0.001504\n",
       "Time:                        16:07:11   Log-Likelihood:                -648.54\n",
       "converged:                       True   LL-Null:                       -649.51\n",
       "Covariance Type:              cluster   LLR p-value:                    0.1622\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      1.4833      0.094     15.726      0.000       1.298       1.668\n",
       "treatment      0.1976      0.157      1.259      0.208      -0.110       0.505\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 473,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('seekhelpfever ~ treatment', ddb[(ddb.week < 20) & ddb.fever4months.notna() & ddb.seekhelpfever.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.272965\n",
      "         Iterations 7\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>seekhelpfever</td>  <th>  No. Observations:  </th>  <td>  1000</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>   996</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     3</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Tue, 02 Mar 2021</td> <th>  Pseudo R-squ.:     </th>  <td>0.01214</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>10:40:52</td>     <th>  Log-Likelihood:    </th> <td> -272.97</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -276.32</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th>  <td>0.08173</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "             <td></td>                <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th>               <td>    2.0565</td> <td>    0.200</td> <td>   10.287</td> <td> 0.000</td> <td>    1.665</td> <td>    2.448</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>pucca[T.True]</th>           <td>    0.4792</td> <td>    0.352</td> <td>    1.360</td> <td> 0.174</td> <td>   -0.211</td> <td>    1.170</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th>               <td>    0.6318</td> <td>    0.348</td> <td>    1.817</td> <td> 0.069</td> <td>   -0.050</td> <td>    1.313</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment:pucca[T.True]</th> <td>   -0.3806</td> <td>    0.540</td> <td>   -0.705</td> <td> 0.481</td> <td>   -1.438</td> <td>    0.677</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   No. Observations:                 1000\n",
       "Model:                          Logit   Df Residuals:                      996\n",
       "Method:                           MLE   Df Model:                            3\n",
       "Date:                Tue, 02 Mar 2021   Pseudo R-squ.:                 0.01214\n",
       "Time:                        10:40:52   Log-Likelihood:                -272.97\n",
       "converged:                       True   LL-Null:                       -276.32\n",
       "Covariance Type:              cluster   LLR p-value:                   0.08173\n",
       "===========================================================================================\n",
       "                              coef    std err          z      P>|z|      [0.025      0.975]\n",
       "-------------------------------------------------------------------------------------------\n",
       "Intercept                   2.0565      0.200     10.287      0.000       1.665       2.448\n",
       "pucca[T.True]               0.4792      0.352      1.360      0.174      -0.211       1.170\n",
       "treatment                   0.6318      0.348      1.817      0.069      -0.050       1.313\n",
       "treatment:pucca[T.True]    -0.3806      0.540     -0.705      0.481      -1.438       0.677\n",
       "===========================================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('seekhelpfever ~ treatment*pucca', ddb[(ddb.week < 10) & ddb.dwelling.notna() & ddb.fever4months.notna() & ddb.seekhelpfever.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.355515\n",
      "         Iterations 6\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>seekhelpfever</td>  <th>  No. Observations:  </th>  <td>   677</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>   671</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     5</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Thu, 18 Feb 2021</td> <th>  Pseudo R-squ.:     </th>  <td>0.01323</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>10:41:37</td>     <th>  Log-Likelihood:    </th> <td> -240.68</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -243.91</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>      <td>cluster</td>     <th>  LLR p-value:       </th>  <td>0.2647</td> \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "                                <td></td>                                   <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th>                                                     <td>    1.4843</td> <td>    0.262</td> <td>    5.672</td> <td> 0.000</td> <td>    0.971</td> <td>    1.997</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>dwelling[T.Pucca (have cement/brick wall and floor]</th>           <td>    0.6776</td> <td>    0.396</td> <td>    1.713</td> <td> 0.087</td> <td>   -0.098</td> <td>    1.453</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>dwelling[T.Semi-pucca]</th>                                        <td>    0.3075</td> <td>    0.415</td> <td>    0.741</td> <td> 0.459</td> <td>   -0.506</td> <td>    1.121</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th>                                                     <td>    0.6492</td> <td>    0.525</td> <td>    1.236</td> <td> 0.217</td> <td>   -0.380</td> <td>    1.679</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment:dwelling[T.Pucca (have cement/brick wall and floor]</th> <td>   -0.4561</td> <td>    0.642</td> <td>   -0.710</td> <td> 0.478</td> <td>   -1.715</td> <td>    0.802</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment:dwelling[T.Semi-pucca]</th>                              <td>   -0.2029</td> <td>    0.727</td> <td>   -0.279</td> <td> 0.780</td> <td>   -1.629</td> <td>    1.223</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:          seekhelpfever   No. Observations:                  677\n",
       "Model:                          Logit   Df Residuals:                      671\n",
       "Method:                           MLE   Df Model:                            5\n",
       "Date:                Thu, 18 Feb 2021   Pseudo R-squ.:                 0.01323\n",
       "Time:                        10:41:37   Log-Likelihood:                -240.68\n",
       "converged:                       True   LL-Null:                       -243.91\n",
       "Covariance Type:              cluster   LLR p-value:                    0.2647\n",
       "=================================================================================================================================\n",
       "                                                                    coef    std err          z      P>|z|      [0.025      0.975]\n",
       "---------------------------------------------------------------------------------------------------------------------------------\n",
       "Intercept                                                         1.4843      0.262      5.672      0.000       0.971       1.997\n",
       "dwelling[T.Pucca (have cement/brick wall and floor]               0.6776      0.396      1.713      0.087      -0.098       1.453\n",
       "dwelling[T.Semi-pucca]                                            0.3075      0.415      0.741      0.459      -0.506       1.121\n",
       "treatment                                                         0.6492      0.525      1.236      0.217      -0.380       1.679\n",
       "treatment:dwelling[T.Pucca (have cement/brick wall and floor]    -0.4561      0.642     -0.710      0.478      -1.715       0.802\n",
       "treatment:dwelling[T.Semi-pucca]                                 -0.2029      0.727     -0.279      0.780      -1.629       1.223\n",
       "=================================================================================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 167,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('seekhelpfever ~ treatment*dwelling', ddb[ddb.dwelling.notna() & ddb.fever4months.notna() & ddb.seekhelpfever.notna()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8145725807070178, 0.8937090377022275, 0.4267786139555295)"
      ]
     },
     "execution_count": 153,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(1.48, 0.6492)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8282108615242946, 0.8628658986241335, 0.20173008263115488)"
      ]
     },
     "execution_count": 136,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(1.573, 0.2663)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Generalized Linear Model Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>   <td>seekhelp_if_fever</td> <th>  No. Observations:  </th>  <td>  1598</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                  <td>GLM</td>        <th>  Df Residuals:      </th>  <td>  1596</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model Family:</th>        <td>Binomial</td>      <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Link Function:</th>         <td>logit</td>       <th>  Scale:             </th> <td>  1.0000</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                <td>IRLS</td>        <th>  Log-Likelihood:    </th> <td> -804.38</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Sun, 07 Feb 2021</td>  <th>  Deviance:          </th> <td>  1516.7</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>18:02:54</td>      <th>  Pearson chi2:      </th> <td>1.76e+03</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Iterations:</th>          <td>4</td>         <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>     <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    1.6309</td> <td>    0.087</td> <td>   18.687</td> <td> 0.000</td> <td>    1.460</td> <td>    1.802</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.1977</td> <td>    0.127</td> <td>    1.561</td> <td> 0.118</td> <td>   -0.050</td> <td>    0.446</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                 Generalized Linear Model Regression Results                  \n",
       "==============================================================================\n",
       "Dep. Variable:      seekhelp_if_fever   No. Observations:                 1598\n",
       "Model:                            GLM   Df Residuals:                     1596\n",
       "Model Family:                Binomial   Df Model:                            1\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -804.38\n",
       "Date:                Sun, 07 Feb 2021   Deviance:                       1516.7\n",
       "Time:                        18:02:54   Pearson chi2:                 1.76e+03\n",
       "No. Iterations:                     4                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      1.6309      0.087     18.687      0.000       1.460       1.802\n",
       "treatment      0.1977      0.127      1.561      0.118      -0.050       0.446\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 119,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m = pd.concat([mat, matb])\n",
    "reg('seekhelp_if_fever ~ treatment', m[m.seekhelp_if_fever.notna()], 'fever2weeks')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8362928924541598, 0.8615948625093831, 0.15455633194263352)"
      ]
     },
     "execution_count": 137,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_percentages(1.6309, 0.1977)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Generalized Linear Model Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>   <td>seekhelp_if_fever</td> <th>  No. Observations:  </th>  <td>  1005</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                  <td>GLM</td>        <th>  Df Residuals:      </th>  <td>  1003</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model Family:</th>        <td>Binomial</td>      <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Link Function:</th>         <td>logit</td>       <th>  Scale:             </th> <td>  1.0000</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                <td>IRLS</td>        <th>  Log-Likelihood:    </th> <td> -556.27</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Sun, 07 Feb 2021</td>  <th>  Deviance:          </th> <td>  1020.5</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>18:20:19</td>      <th>  Pearson chi2:      </th> <td>1.17e+03</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Iterations:</th>          <td>4</td>         <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>     <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>    1.6468</td> <td>    0.107</td> <td>   15.452</td> <td> 0.000</td> <td>    1.438</td> <td>    1.856</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.1449</td> <td>    0.151</td> <td>    0.960</td> <td> 0.337</td> <td>   -0.151</td> <td>    0.441</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                 Generalized Linear Model Regression Results                  \n",
       "==============================================================================\n",
       "Dep. Variable:      seekhelp_if_fever   No. Observations:                 1005\n",
       "Model:                            GLM   Df Residuals:                     1003\n",
       "Model Family:                Binomial   Df Model:                            1\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -556.27\n",
       "Date:                Sun, 07 Feb 2021   Deviance:                       1020.5\n",
       "Time:                        18:20:19   Pearson chi2:                 1.17e+03\n",
       "No. Iterations:                     4                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept      1.6468      0.107     15.452      0.000       1.438       1.856\n",
       "treatment      0.1449      0.151      0.960      0.337      -0.151       0.441\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 137,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('seekhelp_if_fever ~ treatment', mat[mat.seekhelp_if_fever.notna()], 'fever2weeks')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Generalized Linear Model Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>fever4months</td>   <th>  No. Observations:  </th>  <td>  4951</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                  <td>GLM</td>       <th>  Df Residuals:      </th>  <td>  4949</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model Family:</th>        <td>Binomial</td>     <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Link Function:</th>         <td>logit</td>      <th>  Scale:             </th> <td>  1.0000</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                <td>IRLS</td>       <th>  Log-Likelihood:    </th> <td> -1814.3</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Sun, 07 Feb 2021</td> <th>  Deviance:          </th> <td>  3628.7</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>18:12:45</td>     <th>  Pearson chi2:      </th> <td>4.95e+03</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Iterations:</th>          <td>5</td>        <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>    <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>   -1.9782</td> <td>    0.061</td> <td>  -32.593</td> <td> 0.000</td> <td>   -2.097</td> <td>   -1.859</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>   -0.0339</td> <td>    0.088</td> <td>   -0.387</td> <td> 0.699</td> <td>   -0.206</td> <td>    0.138</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                 Generalized Linear Model Regression Results                  \n",
       "==============================================================================\n",
       "Dep. Variable:           fever4months   No. Observations:                 4951\n",
       "Model:                            GLM   Df Residuals:                     4949\n",
       "Model Family:                Binomial   Df Model:                            1\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -1814.3\n",
       "Date:                Sun, 07 Feb 2021   Deviance:                       3628.7\n",
       "Time:                        18:12:45   Pearson chi2:                 4.95e+03\n",
       "No. Iterations:                     5                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept     -1.9782      0.061    -32.593      0.000      -2.097      -1.859\n",
       "treatment     -0.0339      0.088     -0.387      0.699      -0.206       0.138\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('fever4months ~ treatment', matb)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Generalized Linear Model Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>      <td>fever2weeks</td>   <th>  No. Observations:  </th>  <td>  6496</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                  <td>GLM</td>       <th>  Df Residuals:      </th>  <td>  6494</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model Family:</th>        <td>Binomial</td>     <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Link Function:</th>         <td>logit</td>      <th>  Scale:             </th> <td>  1.0000</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                <td>IRLS</td>       <th>  Log-Likelihood:    </th> <td> -4306.4</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Sun, 07 Feb 2021</td> <th>  Deviance:          </th> <td>  5633.2</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>18:12:47</td>     <th>  Pearson chi2:      </th> <td>8.75e+03</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Iterations:</th>          <td>6</td>        <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>    <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>   -2.7039</td> <td>    0.041</td> <td>  -66.740</td> <td> 0.000</td> <td>   -2.783</td> <td>   -2.625</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.1145</td> <td>    0.056</td> <td>    2.041</td> <td> 0.041</td> <td>    0.005</td> <td>    0.224</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                 Generalized Linear Model Regression Results                  \n",
       "==============================================================================\n",
       "Dep. Variable:            fever2weeks   No. Observations:                 6496\n",
       "Model:                            GLM   Df Residuals:                     6494\n",
       "Model Family:                Binomial   Df Model:                            1\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -4306.4\n",
       "Date:                Sun, 07 Feb 2021   Deviance:                       5633.2\n",
       "Time:                        18:12:47   Pearson chi2:                 8.75e+03\n",
       "No. Iterations:                     6                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept     -2.7039      0.041    -66.740      0.000      -2.783      -2.625\n",
       "treatment      0.1145      0.056      2.041      0.041       0.005       0.224\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 126,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('fever2weeks ~ treatment', mat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Generalized Linear Model Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>    <td>malaria4months</td>  <th>  No. Observations:  </th>  <td>  5001</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                  <td>GLM</td>       <th>  Df Residuals:      </th>  <td>  4999</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model Family:</th>        <td>Binomial</td>     <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Link Function:</th>         <td>logit</td>      <th>  Scale:             </th> <td>  1.0000</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                <td>IRLS</td>       <th>  Log-Likelihood:    </th> <td> -965.49</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Sun, 07 Feb 2021</td> <th>  Deviance:          </th> <td>  1931.0</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>17:59:59</td>     <th>  Pearson chi2:      </th> <td>5.00e+03</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Iterations:</th>          <td>6</td>        <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>    <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>   -3.0466</td> <td>    0.095</td> <td>  -32.197</td> <td> 0.000</td> <td>   -3.232</td> <td>   -2.861</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>    0.1270</td> <td>    0.132</td> <td>    0.962</td> <td> 0.336</td> <td>   -0.132</td> <td>    0.386</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                 Generalized Linear Model Regression Results                  \n",
       "==============================================================================\n",
       "Dep. Variable:         malaria4months   No. Observations:                 5001\n",
       "Model:                            GLM   Df Residuals:                     4999\n",
       "Model Family:                Binomial   Df Model:                            1\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -965.49\n",
       "Date:                Sun, 07 Feb 2021   Deviance:                       1931.0\n",
       "Time:                        17:59:59   Pearson chi2:                 5.00e+03\n",
       "No. Iterations:                     6                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept     -3.0466      0.095    -32.197      0.000      -3.232      -2.861\n",
       "treatment      0.1270      0.132      0.962      0.336      -0.132       0.386\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('malaria4months ~ treatment', matb)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.045364488367139494, 0.5317073940719488)"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(-3.0466), expit(0.1270)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Generalized Linear Model Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>     <td>malaria2weeks</td>  <th>  No. Observations:  </th>  <td>  6496</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                  <td>GLM</td>       <th>  Df Residuals:      </th>  <td>  6494</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model Family:</th>        <td>Binomial</td>     <th>  Df Model:          </th>  <td>     1</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Link Function:</th>         <td>logit</td>      <th>  Scale:             </th> <td>  1.0000</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                <td>IRLS</td>       <th>  Log-Likelihood:    </th> <td> -1526.8</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Sun, 07 Feb 2021</td> <th>  Deviance:          </th> <td>  3053.5</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>18:15:45</td>     <th>  Pearson chi2:      </th> <td>6.50e+03</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>No. Iterations:</th>          <td>6</td>        <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>    <th>                     </th>     <td> </td>   \n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "      <td></td>         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th> <td>   -2.6843</td> <td>    0.072</td> <td>  -37.365</td> <td> 0.000</td> <td>   -2.825</td> <td>   -2.543</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th> <td>   -0.0320</td> <td>    0.102</td> <td>   -0.313</td> <td> 0.754</td> <td>   -0.232</td> <td>    0.168</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                 Generalized Linear Model Regression Results                  \n",
       "==============================================================================\n",
       "Dep. Variable:          malaria2weeks   No. Observations:                 6496\n",
       "Model:                            GLM   Df Residuals:                     6494\n",
       "Model Family:                Binomial   Df Model:                            1\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -1526.8\n",
       "Date:                Sun, 07 Feb 2021   Deviance:                       3053.5\n",
       "Time:                        18:15:45   Pearson chi2:                 6.50e+03\n",
       "No. Iterations:                     6                                         \n",
       "Covariance Type:            nonrobust                                         \n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "------------------------------------------------------------------------------\n",
       "Intercept     -2.6843      0.072    -37.365      0.000      -2.825      -2.543\n",
       "treatment     -0.0320      0.102     -0.313      0.754      -0.232       0.168\n",
       "==============================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reg('malaria2weeks ~ treatment', mat, 'foo')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.06390615823520344, 0.061439175217968434)"
      ]
     },
     "execution_count": 132,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(-2.6843), expit(-2.6943 - 0.0320)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dd[dd.wave == 0].groupby('treatment').long_sleeves.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('as_type(under_net, int) ~ treatment:as_type(wave, str) + avg_under_net', dd[(dd.userid.isin(mat[mat.net_ownership].userid)) & (dd.wave > 0)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('as_type(under_net, int) ~ treatment:as_type(month, str) + avg_under_net', dd[(dd.userid.isin(mat[mat.net_ownership].userid)) & (dd.wave > 0)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('as_type(long_sleeves, int) ~ treatment:as_type(month, str) + avg_long_sleeves', dd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('as_type(long_sleeves, int) ~ treatment', dd[dd.wave == 0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('as_type(under_net, int) ~ treatment', dd[dd.wave == 0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('as_type(malaria_now, int) ~ treatment', dd[dd.wave == 0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('malaria_now ~ treatment', mat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "mat['num_under_net'].max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('num_under_net ~ treatment', mat[(mat['owns_net'] == 'Yes')]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('long_sleeves ~ treatment + avg_long_sleeves', mat) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('under_net ~ treatment', mat[mat['owns_net'] == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('under_net ~ treatment*pregnant_in_house', mat[mat['owns_net'] == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('under_net ~ treatment*quantile(avg_malaria_now, 0.5)', mat[mat['owns_net'] == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('under_net ~ treatment*kutcha*quantile(avg_malaria_now, 0.5)', mat[mat['owns_net'] == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1413,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6454503734274153"
      ]
     },
     "execution_count": 1413,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.5991)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1412,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6506960099929763"
      ]
     },
     "execution_count": 1412,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.5991 + 0.023)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1416,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5891372593292695"
      ]
     },
     "execution_count": 1416,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.5991 - 0.2387)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1414,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6316262977304613"
      ]
     },
     "execution_count": 1414,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.5991 + 0.023 + 0.1558 - 0.2387)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('under_net ~ treatment', mat[mat['owns_net'] == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1378,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1379,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6292498476593299"
      ]
     },
     "execution_count": 1379,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.529)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1388,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.617747874769249"
      ]
     },
     "execution_count": 1388,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.48)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1389,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6410674063348171"
      ]
     },
     "execution_count": 1389,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.58)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1405,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.5863750605361624, 0.648168881838421)"
      ]
     },
     "execution_count": 1405,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.349), expit(0.611)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1404,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.6201064323430902, 0.7045375762042413)"
      ]
     },
     "execution_count": 1404,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expit(0.58 - 0.09), expit(0.58 + 0.289)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1402,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7045375762042413"
      ]
     },
     "execution_count": 1402,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('under_net ~ treatment*lower_caste*quantile(avg_malaria_now, 0.5) + avg_under_net', mat[mat['owns_net'] == 'Yes'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "reg('long_sleeves ~ treatment*lower_caste*quantile(avg_malaria_now, 0.5) + avg_long_sleeves', mat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 415,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 415,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mat.malaria_now.sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 410,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])"
      ]
     },
     "execution_count": 410,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mat.groupby('clusterid').malaria_now.sum().values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 383,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  malaria_now\n",
       "0          0              0.917864\n",
       "           1              0.082136\n",
       "1          0              0.910769\n",
       "           1              0.089231\n",
       "Name: malaria_now, dtype: float64"
      ]
     },
     "execution_count": 383,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mat.groupby('treatment').malaria_now.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df.clusterid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df.wave == 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 768,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dist_info['high_malaria_district'] = dist_info.avg_malaria_now > 0.02"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 495,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 2, 1, 0, 0, 1, 0,\n",
       "       0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1,\n",
       "       0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,\n",
       "       2, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0])"
      ]
     },
     "execution_count": 495,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dd[dd.wave == 0].groupby('clusterid').apply(lambda df: df.malaria_now.sum()).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dist_info"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "mm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 543,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.287730\n",
      "         Iterations 8\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<table class=\"simpletable\">\n",
       "<caption>Logit Regression Results</caption>\n",
       "<tr>\n",
       "  <th>Dep. Variable:</th>      <td>malaria_now</td>   <th>  No. Observations:  </th>  <td>  1949</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Model:</th>                 <td>Logit</td>      <th>  Df Residuals:      </th>  <td>  1941</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Method:</th>                 <td>MLE</td>       <th>  Df Model:          </th>  <td>     7</td> \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Date:</th>            <td>Wed, 02 Dec 2020</td> <th>  Pseudo R-squ.:     </th>  <td>0.01610</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Time:</th>                <td>16:37:02</td>     <th>  Log-Likelihood:    </th> <td> -560.79</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>converged:</th>             <td>True</td>       <th>  LL-Null:           </th> <td> -569.96</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Covariance Type:</th>     <td>nonrobust</td>    <th>  LLR p-value:       </th>  <td>0.01047</td>\n",
       "</tr>\n",
       "</table>\n",
       "<table class=\"simpletable\">\n",
       "<tr>\n",
       "                       <td></td>                         <th>coef</th>     <th>std err</th>      <th>z</th>      <th>P>|z|</th>  <th>[0.025</th>    <th>0.975]</th>  \n",
       "</tr>\n",
       "<tr>\n",
       "  <th>Intercept</th>                                  <td>   -1.9847</td> <td>    0.738</td> <td>   -2.689</td> <td> 0.007</td> <td>   -3.431</td> <td>   -0.538</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment</th>                                  <td>    0.4164</td> <td>    1.093</td> <td>    0.381</td> <td> 0.703</td> <td>   -1.725</td> <td>    2.558</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>avg_malaria_now</th>                            <td>    0.0445</td> <td>   38.359</td> <td>    0.001</td> <td> 0.999</td> <td>  -75.138</td> <td>   75.227</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment:avg_malaria_now</th>                  <td>  -11.6342</td> <td>   51.275</td> <td>   -0.227</td> <td> 0.821</td> <td> -112.130</td> <td>   88.862</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>cost_per_message</th>                           <td>   -0.0403</td> <td>    0.034</td> <td>   -1.195</td> <td> 0.232</td> <td>   -0.106</td> <td>    0.026</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment:cost_per_message</th>                 <td>   -0.0314</td> <td>    0.054</td> <td>   -0.581</td> <td> 0.561</td> <td>   -0.138</td> <td>    0.075</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>avg_malaria_now:cost_per_message</th>           <td>    0.9171</td> <td>    1.603</td> <td>    0.572</td> <td> 0.567</td> <td>   -2.225</td> <td>    4.059</td>\n",
       "</tr>\n",
       "<tr>\n",
       "  <th>treatment:avg_malaria_now:cost_per_message</th> <td>    0.9822</td> <td>    2.285</td> <td>    0.430</td> <td> 0.667</td> <td>   -3.496</td> <td>    5.461</td>\n",
       "</tr>\n",
       "</table>"
      ],
      "text/plain": [
       "<class 'statsmodels.iolib.summary.Summary'>\n",
       "\"\"\"\n",
       "                           Logit Regression Results                           \n",
       "==============================================================================\n",
       "Dep. Variable:            malaria_now   No. Observations:                 1949\n",
       "Model:                          Logit   Df Residuals:                     1941\n",
       "Method:                           MLE   Df Model:                            7\n",
       "Date:                Wed, 02 Dec 2020   Pseudo R-squ.:                 0.01610\n",
       "Time:                        16:37:02   Log-Likelihood:                -560.79\n",
       "converged:                       True   LL-Null:                       -569.96\n",
       "Covariance Type:            nonrobust   LLR p-value:                   0.01047\n",
       "==============================================================================================================\n",
       "                                                 coef    std err          z      P>|z|      [0.025      0.975]\n",
       "--------------------------------------------------------------------------------------------------------------\n",
       "Intercept                                     -1.9847      0.738     -2.689      0.007      -3.431      -0.538\n",
       "treatment                                      0.4164      1.093      0.381      0.703      -1.725       2.558\n",
       "avg_malaria_now                                0.0445     38.359      0.001      0.999     -75.138      75.227\n",
       "treatment:avg_malaria_now                    -11.6342     51.275     -0.227      0.821    -112.130      88.862\n",
       "cost_per_message                              -0.0403      0.034     -1.195      0.232      -0.106       0.026\n",
       "treatment:cost_per_message                    -0.0314      0.054     -0.581      0.561      -0.138       0.075\n",
       "avg_malaria_now:cost_per_message               0.9171      1.603      0.572      0.567      -2.225       4.059\n",
       "treatment:avg_malaria_now:cost_per_message     0.9822      2.285      0.430      0.667      -3.496       5.461\n",
       "==============================================================================================================\n",
       "\"\"\""
      ]
     },
     "execution_count": 543,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smf.logit('malaria_now ~ treatment*avg_malaria_now*cost_per_message', mm).fit().summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 438,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  translated_response\n",
       "0          Yes                    0.569815\n",
       "           No                     0.430185\n",
       "1          Yes                    0.562628\n",
       "           No                     0.437372\n",
       "Name: translated_response, dtype: float64"
      ]
     },
     "execution_count": 438,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[(df.variable == 'net_lastnight') & (df.wave == 0)].groupby('treatment').translated_response.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 440,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1197     2020-07-23 08:06:51.476000+00:00\n",
       "1190     2020-07-23 08:07:00.147000+00:00\n",
       "1165     2020-07-23 08:07:55.339000+00:00\n",
       "1181     2020-07-23 08:08:15.220000+00:00\n",
       "1196     2020-07-23 08:08:21.454000+00:00\n",
       "                       ...               \n",
       "119064   2020-12-02 08:32:00.924000+00:00\n",
       "95367    2020-12-02 08:57:51.108000+00:00\n",
       "95615    2020-12-02 08:58:27.031000+00:00\n",
       "95616    2020-12-02 08:59:03.112000+00:00\n",
       "95623    2020-12-02 08:59:11.999000+00:00\n",
       "Name: timestamp, Length: 83171, dtype: datetime64[ns, psycopg2.tz.FixedOffsetTimezone(offset=0, name=None)]"
      ]
     },
     "execution_count": 440,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.timestamp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 439,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  translated_response\n",
       "0          Yes                    0.519062\n",
       "           No                     0.480938\n",
       "1          Yes                    0.572917\n",
       "           No                     0.427083\n",
       "Name: translated_response, dtype: float64"
      ]
     },
     "execution_count": 439,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[(df.variable == 'net_lastnight') & (df.wave > 5)].groupby('treatment').translated_response.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment\n",
       "0.0    3.153181\n",
       "1.0    3.222683\n",
       "Name: summed, dtype: float64"
      ]
     },
     "execution_count": 198,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[(df.variable == 'net_lastnight') & (df.wave > 0)] \\\n",
    "    .groupby('userid') \\\n",
    "    .apply(lambda df: df.assign(summed = (df.translated_response == 'Yes').sum())) \\\n",
    "    .groupby('treatment') \\\n",
    "    .summed.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df[(df.variable == 'long_sleeves') & (df.wave == 0)].groupby('treatment').translated_response.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 408,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "treatment  translated_response\n",
       "0          Yes                    0.675000\n",
       "           No                     0.325000\n",
       "1          Yes                    0.653153\n",
       "           No                     0.346847\n",
       "Name: translated_response, dtype: float64"
      ]
     },
     "execution_count": 408,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[(df.variable == 'long_sleeves') & (df.wave > 5)].groupby('treatment').translated_response.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1070,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "bl = dda[dda.wave == 0]\n",
    "bl = binarize_cols(bl, [('malaria5year', 'Yes')])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1073,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1859.0"
      ]
     },
     "execution_count": 1073,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bl.malaria5year"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1075,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "164.0"
      ]
     },
     "execution_count": 1075,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 1090,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def _convert_cat(col):\n",
    "    try:\n",
    "        return col.cat.codes\n",
    "    except AttributeError:\n",
    "        return None\n",
    "\n",
    "def convert_cats(df):\n",
    "    cols = [(col, _convert_cat(df[col])) for col in df.columns]\n",
    "    cols =  [(n, c) for n, c in cols if c is not None]\n",
    "    names, cols = zip(*cols)\n",
    "    return cols\n",
    "\n",
    "def ordinalize(df, col, vals):\n",
    "    t = CategoricalDtype(categories=vals, ordered=True)\n",
    "    df[col] = df[col].astype(t)\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1101,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "bl = ordinalize(bl, 'distancemedicalcenter', ['Less than 15 minutes', 'Between 15 and 30 minutes', 'Between 30 and 60 minutes', 'More than 60 minutes'])\n",
    "bl = ordinalize(bl, 'education', ['Never attended school', 'Primary', 'Secondary', 'University degree or higher'])\n",
    "bl = ordinalize(bl, 'home', ['Kutcha (made of mud, tin, straw)', 'Semi-pucca', 'Pucca (have cement/brick wall and floor'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1265,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "bl['age'] = bl.age.map(lambda x: cast_number(15, 90, x)).astype(float)\n",
    "bl['familymembers'] = bl.familymembers.map(lambda x: cast_number(0, 30, x)).astype(float)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1253,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "bl['unemployed'] = bl.occupation == 'Unemployed'\n",
    "bl['self_employed'] = bl.occupation == 'Self-employed'\n",
    "bl['student'] = bl.occupation == 'Student'\n",
    "bl['employed_fulltime'] = bl.occupation == 'Employed-full time'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1258,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "bl['general_caste'] = bl.caste == 'General'\n",
    "bl['lower_caste'] = bl.caste.isin(['OBC', 'SC/Dalit', 'ST'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1318,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "bl['female'] = bl.gender == 'Female'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1337,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier, DecisionTreeRegressor\n",
    "from dtreeviz.trees import *\n",
    "\n",
    "model = DecisionTreeClassifier(max_depth=2)\n",
    "\n",
    "df = bl.dropna(subset=['malaria2weeks', 'distancemedicalcenter', 'education', 'home', 'familymembers', 'age']).reset_index(drop=True)\n",
    "df = df[df.age >= 18].reset_index(drop=True)\n",
    "y = df.malaria2weeks\n",
    "# df['kutcha'] = df.home == 'Kutcha (made of mud, tin, straw)'\n",
    "# df['pucca'] = df.home == 'Pucca (have cement/brick wall and floor'\n",
    "picked = ['age', 'female', 'unemployed', 'self_employed', 'student', 'employed_fulltime', 'general_caste', 'lower_caste']\n",
    "X = df[picked]\n",
    "\n",
    "X = pd.concat([X, *convert_cats(df[['distancemedicalcenter', 'education', 'home']])], axis=1)\n",
    "X.columns = picked + ['distancemedicalcenter', 'education', 'dwelling']\n",
    "\n",
    "model.fit(X, y)\n",
    "\n",
    "viz = dtreeviz(model, \n",
    "               X, \n",
    "               y,\n",
    "               target_name='malaria2weeks',\n",
    "               feature_names=X.columns, \n",
    "               # fancy=False\n",
    "              )  \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "viz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1333,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "viz.save('outs/images/malaria5year.svg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1334,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dwelling                -0.068\n",
       "general_caste           -0.048\n",
       "self_employed           -0.032\n",
       "age                     -0.023\n",
       "female                  -0.012\n",
       "employed_fulltime       -0.009\n",
       "student                  0.009\n",
       "familymembers            0.017\n",
       "unemployed               0.021\n",
       "education                0.038\n",
       "distancemedicalcenter    0.049\n",
       "lower_caste              0.056\n",
       "malaria2weeks            1.000\n",
       "Name: malaria2weeks, dtype: float64"
      ]
     },
     "execution_count": 1334,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X['malaria2weeks'] = df.malaria5year\n",
    "X.corr()['malaria2weeks'].round(3).sort_values()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1385,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ddb['kutcha'] = ddb.dwelling == 'Kutcha (made of mud, tin, straw)'\n",
    "ddb['pucca'] = ddb.dwelling == 'Pucca (have cement/brick wall and floor'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "with sns.color_palette('Set2'):\n",
    "    # _, ax = plt.subplots(figsize=(16, 5))\n",
    "    sns.displot(ddb[ddb.dwelling.notna()].groupby('stratumid').kutcha.mean(), alpha = 0.5, kde=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1220,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "X['y'] = y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "X.corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# LOOK AT CHANGE IN INDIVIDUAL BEHAVIOR, DIF IN DIF!\n",
    "# Too weak, 82 users :(\n",
    "\n",
    "early_fever_users = dda[(dda.week < 36) & dda.seekhelpfever.notna()].userid.unique()\n",
    "fu = dda[(dda.userid.isin(early_fever_users)) & (dda.week > 38) & dda.seekhelpfever.notna()].groupby('userid').apply(lambda df: df.sort_values('week').iloc[-1]).reset_index(drop=True)\n",
    "\n",
    "base = dda[(dda.userid.isin(fu.userid)) & (dda.week < 36) & dda.seekhelpfever.notna()].groupby('userid').apply(lambda df: df.sort_values('week').iloc[0]).reset_index(drop=True).rename(columns = {'seekhelpfever': 'pre_seekhelpfever'})\n",
    "\n",
    "combo = base[['userid', 'pre_seekhelpfever', 'treatment']].merge(fu[['userid', 'seekhelpfever']])\n",
    "combo['dif'] = combo.seekhelpfever - combo.pre_seekhelpfever\n",
    "\n",
    "smf.ols('dif ~ treatment', combo).fit().summary()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "argv": [
    "python",
    "-m",
    "ipykernel_launcher",
    "-f",
    "{connection_file}"
   ],
   "display_name": "Python 3",
   "env": null,
   "interrupt_mode": "signal",
   "language": "python",
   "metadata": null,
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.0"
  },
  "name": "responses.ipynb"
 },
 "nbformat": 4,
 "nbformat_minor": 4
}