newBeta12.R

options(scipen = 25, max.print = 10000000)
library(jsonlite)
library(plyr)
library(dplyr)
library(tidyr)
library(ggplot2)
library(TTR)
library(zoo)
library(jtools)


#########################################################################################################################
#   What I think I did was in a note from Cynthia.  I have no confidence whether I did this correctly or not            #
#########################################################################################################################
# Start on the day of 100 confirmed, drop countries with less than 100 confirmed or less than 10 dead
# Regressions should go through origin                                                                                  #
#                                                                                                                       #
# calculate log of difference: log(7dMA at time T - 7dMA at time t-1) vs. log of cumulative confirmed (x-axis)          #
# calculate difference of log: log of 7dMA at time T - log of 7dMA at time T-1 vs. log of cumulative confirmed (x-axis) #
# raw difference: 7dMA at time T - 7dMA at time T-1 vs. total cumulative confirmed                                      #
#########################################################################################################################
# Added Bootstrap from Roy #
############################


#########################################################################################
### Funtion to create a regression plot and annotate the Y intercept, P, and Formula ###
#########################################################################################


#########################################################################################
### Funtions from Roy to do the bootstrapping                                         ###
#########################################################################################

lm_boot <-
  function(data, ind) {
    lm(log_of_difference ~   0 + log_cum_confirmed, data = data[ind, ])$coef
  }

rsq_boot <- function(data, ind) {
  fit <-
    lm(log_of_difference ~   0 + log_cum_confirmed, data = data[ind, ])
  return(summary(fit)$r.square)
}

#########################################################################################
### Function to plot regression                                                       ###
#########################################################################################

ggplotRegression <- function(fit, country) {
  ggplot(fit$model, aes_string(x = names(fit$model)[2], y = names(fit$model)[1])) +
    geom_point() +
    stat_smooth(method = "lm", col = "red") +
    labs(
      title = paste(
        " Country =",
        country,
        "Adj R2 = ",
        signif(summary(fit)$adj.r.squared, 5),
        "Coefficient =",
        signif(fit$coef[[1]], 5)
#         " P =",
# #        signif(summary(fit)$coef[2, 4], 5),
#         " N = ", nrow(fit$model)
    )
    )

}

#########################################################################################
### Function to output all plots                                                      ###
#########################################################################################

plot.countries <- function(bigtable = C19DF) {
  if (dir.exists("output")) {
    
  }
  else {
    dir.create("output")
  }
  countries <- unique(bigtable$country)
  for (i in seq_along(countries)) {
    item <- C19DF[(C19DF$country == countries[i]), ]
    
    plot <-
      ggplotRegression(lm(ln_total_cases ~ 0 + ln_ma_cases, data = item),
                       countries[i])
    
    ggsave(
      filename = paste0("output/",
                        countries[i],
                        "_xy_plot.png"),
      plot = plot,
      width = 11,
      height = 8.5,
      units = "in"
    )
    print(plot)
  }
}


#########################################################################################
###           PROGRAM START                                                           ###
#########################################################################################




#
# The following lines read the JHS data for all countries  - I just grab it all, but easy to take a subset
#
base_url <-
  "https://pomber.github.io/covid19/timeseries.json" # REST API for JHS data


if (!file.exists("timeseries.json")) {

  download.file(base_url,
                "timeseries.json",
                "wget")
}

c19JSON <-
  fromJSON(txt = "timeseries.json") # Get all country, all date, confirmed, deaths, recovered


C19DF <-
  do.call(rbind, unname(Map(cbind, country = names(c19JSON), c19JSON)))
C19DF$country <- as.character(C19DF$country)
C19DF <- C19DF[order(C19DF$country), ]

############  At this point C19DF is a dataframe with all JHS country data

#######  DATAFRAME ####
#  country        - country
#  date           - date of reporting
#  confirmed      - total reported cases
#  deaths         - deaths
#  recovered      - recovered
#  new_cases      - first difference of total reported cases
#  ma_cases       - 7 day moving average of total cases
#  new_ma_cases   - period (daily) change in 7 day ma
#  ln_total_cases - ln of total cases
#  ln_new_cases   - ln of new cases
#  ln_ma_cases    - ln of movin average of total cases
#  ln_new_ma_cases- (ln of moving average - ln of previous (daily) moving average)

cnames <-
  c(
    "country",
    "date",
    "total_cases",
    "deaths",
    "recovered",
    "new_cases",
    "ma_cases",
    "new_ma_cases",
    "ln_total_cases",
    "ln_new_cases",
    "ln_ma_cases",
    "ln_new_ma_cases",
    "diff_ln_ma_cases",
    "new_deaths",
    "ln_deaths",
    "ln_new_deaths"
  )

options(warn = -1)

C19DF <- C19DF[!(C19DF$country == "Diamond Princess"), ]

# ----- Next line to simplify table for debugging -----"
# mask <- startsWith(as.character(C19DF$country),"U")
# C19DF <- C19DF[mask,]
#------------------------------------------------------

C19DF <-
  C19DF %>% group_by(country) %>% mutate(new_cases = confirmed - lag(confirmed))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(ma_cases = runMean(confirmed, 7))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(new_ma_cases = ma_cases - lag(ma_cases))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(ln_total_cases = log(confirmed))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(ln_new_cases = log(new_cases))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(ln_ma_cases = log(ma_cases))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(ln_new_ma_cases = log(new_ma_cases))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(diff_ln_ma_cases = ln_ma_cases - lag(ln_ma_cases))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(new_deaths = deaths - lag(deaths))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(ln_deaths = log(deaths))
C19DF <-
  C19DF %>% group_by(country) %>% mutate(ln_new_deaths = log(new_deaths))




colnames(C19DF) <- cnames

C19DF <-
  do.call(data.frame, lapply(C19DF, function(x)
    replace(x, is.infinite(x), NA)))
C19DF <-
  subset(C19DF, total_cases >= 100 &
           deaths > 10)       # Should it be 100 OR 10?????
D1 <- table(C19DF$country)                                     #
C19DF <-
  C19DF[C19DF$country %in% names(D1[D1 >= 25]),]       # AND at least 25 observations

countries <- as.character(unique(C19DF$country))




plot.countries(C19DF)