examples.Rmd

---
title: "R Notebook"
output: html_notebook
---

```{r}
library(tidyverse)
library(GGally)
library(tidytext)
library(wordcloud2) 
library(topicmodels)
```

| Field | Question |
| :- | :- |
| timestamp | Timestamp |
| q1 | How many hours did you sleep last night? |
| q2 | How many hours did you work (e.g. do homework, or a job) after you left school? |
| **q3** **| How many hours did you spend relaxing after you left school?** |
| q4 | What did you do to relax last night? |
| q5 | How do you feel today? |
| q6 | Which best describes you (your answer should be the same every day) |


## Start by labelling your data

- Labeling help you understand your data
- Folks often overestimate how long it takes
- If outsourcing...
  - Start by labeling the data yourself
  - Have an odd number of people labeling
  - Come up with consistent rules (heuristics)

```{r}
# Use set_names to name columns

survey_df <- read_csv("survey_results.csv") %>%
  set_names(
    c(
      "timestamp",
      "q1",
      "q2", 
      "q3",
      "q4",
      "q5",
      "q6",
      "watch_shows_movies",
      "sleep",
      "chat",
      "food",
      "friends",
      "video_games",
      "listen_music",
      "book"
    )
  )
survey_df
```

## Plot Bar Chart of Common Tags

```{r}
# Select label columns
# Use colSums to get sum of each column
# Use stack to get "long" format of column sums

survey_df %>%
  select(
    watch_shows_movies,
    sleep,
    chat,
    food,
    friends,
    video_games,
    listen_music,
    book
  ) %>%
  colSums() %>%
  stack() %>%
  ggplot() +
    geom_bar(aes(x = values, y = ind), stat = "identity")
```

## Plot Correlation Between Tags

```{r}
# Use select and ggcorr to plot correlations
survey_df %>%
  select(
    watch_shows_movies,
    sleep,
    chat,
    food,
    friends,
    video_games,
    listen_music,
    book
  ) %>%
  ggcorr()
```

## Plot Word Counts

Each **document** is a response, and each **token** is a word.

```{r}
# select the q4 column
# unnest_tokens separates each documents into tokens
# count() the occurrences of each token
# set the width and height while ggsave
survey_df %>%
  select(q4) %>%
  unnest_tokens(word, q4) %>%
  count(word, sort = TRUE) %>%
  ggplot() +
    geom_bar(aes(y = reorder(word, n), x = n), stat = "identity")

ggsave(
  filename = "word_counts.png",
  width = 5,
  height = 10
)
```

## Word Clouds (and why they not as good)
https://r-graph-gallery.com/196-the-wordcloud2-library.html
Warning: https://www.data-to-viz.com/graph/wordcloud.html

```{r}
survey_df %>%
  select(q4) %>%
  unnest_tokens(word, q4) %>%
  count(word) %>%
  wordcloud2()
```

## Topic Modelling
https://www.tidytextmining.com/topicmodeling.html

```{r}
survey_lda <- survey_df %>%
  select(q4) %>%
  mutate(document = 1:n()) %>%
  unnest_tokens(word, q4) %>%
  group_by(document) %>%
  add_count(word) %>%
  cast_dtm(term = word, document = document, value = n) %>%
  LDA(k = 2, control = list(seed = 1234))

survey_topics <- tidy(survey_lda, matrix = "beta")
survey_topics
```

```{r}
survey_top_terms <- survey_topics %>%
  group_by(topic) %>%
  slice_max(beta, n = 10) %>% 
  ungroup() %>%
  arrange(topic, -beta)

survey_top_terms %>%
  mutate(term = reorder_within(term, beta, topic)) %>%
  ggplot(aes(beta, term, fill = factor(topic))) +
  geom_col(show.legend = FALSE) +
  facet_wrap(~ topic, scales = "free") +
  scale_y_reordered()
```


```{r}
beta_wide <- survey_topics %>%
  mutate(topic = paste0("topic", topic)) %>%
  pivot_wider(names_from = topic, values_from = beta) %>% 
  filter(topic1 > .001 | topic2 > .001) %>%
  mutate(log_ratio = log2(topic2 / topic1))

beta_wide %>%
  ggplot() +
  geom_bar(aes(y = reorder(term, log_ratio), x = log_ratio), stat = "identity")

ggsave(
  filename = "two_topic_model.png",
  width = 5,
  height = 10
)
```