2021_03_23_un_votes.Rmd

---
title: "TidyTemplate"
date: 2021-03-23
output: html_output
---


```{r setup, include=FALSE}

knitr::opts_chunk$set(echo = TRUE)

library(tidyverse)
library(tidytuesdayR)
library(scales)
theme_set(theme_light())

```

```{r Load}
tt <- tt_load("2021-03-23")
```

```{r}
unvotes <- tt$unvotes %>%
  mutate(vote_number = match(vote, c("no", "abstain", "yes")) - 2) %>%
  left_join(tt$roll_calls %>%
              select(rcid, date, amend), by = "rcid")

unvotes %>%
  count(country, sort = TRUE)
```

```{r}
summarize_votes <- function(tbl, min_votes = 10) {
  tbl %>%
    summarize(n_votes = n(),
              n_yes = sum(vote == "yes"),
              pct_yes = n_yes / n_votes,
              .groups = "drop") %>%
  filter(n_votes >= min_votes) %>%
  arrange(desc(pct_yes))
}

by_country <- unvotes %>%
  group_by(country, country_code) %>%
  summarize_votes()

by_country %>%
  slice(c(1:10, (n() - 10):n())) %>%
  mutate(country = fct_reorder(country, pct_yes)) %>%
  ggplot(aes(pct_yes, country)) +
  geom_point(aes(size = n_votes)) +
  scale_x_continuous(labels = percent) +
  labs(x = "% of yes votes in UN",
       title = "What countries voted yes the least")
```

```{r}
library(lubridate)
unvote

by_year <- unvotes %>%
  group_by(year = year(date)) %>%
  summarize_votes()

by_year %>%
  ggplot(aes(year, pct_yes)) +
  geom_line() +
  expand_limits(y = 0)

by_country_year <- unvotes %>%
  bind_rows(unvotes %>% mutate(country = "Overall")) %>%
  group_by(year = year(date), country, country_code) %>%
  summarize_votes()

by_country_year %>%
  filter(country %in% c("United States", "Canada", "Mali", "Israel")) %>%
  mutate(country = fct_reorder(country, pct_yes)) %>%
  ggplot(aes(year, pct_yes)) +
  geom_line(data = by_year, size = 2) +
  geom_line(aes(color = country)) +
  scale_y_continuous(labels = percent) +
  scale_color_discrete(guide = guide_legend(reverse = TRUE)) +
  expand_limits(y = 0) +
  labs(y = "% yes votes")

by_country_year %>%
  filter(country %in% c("United States", "Canada", "Mali", "Israel",
                        "Germany", "France")) %>%
  mutate(country = fct_reorder(country, pct_yes)) %>%
  ggplot(aes(year, pct_yes, color = country)) +
  geom_line() +
  scale_y_continuous(labels = percent) +
  expand_limits(y = 0) +
  facet_wrap(~ country) +
  theme(legend.position = "none") +
  labs(y = "% yes votes")
```

World map

```{r}
library(ggthemes)
library(fuzzyjoin)

world_data <- map_data("world") %>%
  as_tibble() %>%
  regex_left_join(maps::iso3166 %>%
                    select(mapname, country_code = a2),
                  c(region = "mapname")) %>%
  filter(region != "Antarctica")

world_data %>%
  left_join(by_country, by = "country_code") %>%
  ggplot(aes(long, lat, group = group, fill = pct_yes)) +
  geom_polygon() +
  theme_map() +
  scale_fill_gradient2(low = "red",
                       high = "blue",
                       midpoint = .6,
                       labels = percent) +
  labs(fill = "% yes votes")
```

By continent

```{r}
library(countrycode)

plot_by <- function(tbl, category) {
  tbl %>%
    filter(!is.na({{ category }})) %>%
    mutate(category = fct_reorder({{ category }}, pct_yes)) %>%
    ggplot(aes(year, pct_yes)) +
    geom_line(aes(color = category)) +
    scale_y_continuous(labels = percent) +
    scale_color_discrete(guide = guide_legend(reverse = TRUE)) +
    expand_limits(y = 0) +
    labs(y = "% yes votes",
         x = "Year")
}

unvotes %>%
  group_by(continent = countrycode(country_code, "iso2c", "continent"),
           year = year(date)) %>%
  summarize_votes() %>%
  plot_by(continent)
```

```{r}
library(WDI)

country_incomes <- WDI(indicator = c(gdp_per_capita = "NY.GDP.PCAP.PP.KD",
                                     pop = "SP.POP.TOTL"),
                       start = 2005, end = 2005, extra = TRUE) %>%
  as_tibble() %>%
  select(country_code = iso2c, income, gdp_per_capita, pop) %>%
  filter(!is.na(income)) %>%
  mutate(income = fct_relevel(income, "Low income", "Lower middle income", "Upper middle income"))

unvotes %>%
  inner_join(country_incomes, by = "country_code") %>%
  group_by(income,
           year = year(date)) %>%
  summarize_votes() %>%
  plot_by(income)
```

### Correlation between countries

```{r}
unvotes %>%
  filter(country %in% c("India", "Canada")) %>%
  select(rcid, country, vote_number) %>%
  spread(country, vote_number, fill = 0) %>%
  summarize(correlation = cor(Canada, India))

library(widyr)

country_correlations <- unvotes %>%
  filter(country != "Zanzibar") %>%
  pairwise_cor(country, rcid, vote_number, sort = TRUE) %>%
  mutate(continent1 = countrycode(item1, "country.name", "continent"),
         continent2 = countrycode(item2, "country.name", "continent"))

country_correlations %>%
  filter(item1 == "South Africa") %>%
  slice(c(1:16, n() - 0:15)) %>%
  mutate(country = fct_reorder(item2, correlation)) %>%
  ggplot(aes(correlation, country)) +
  geom_errorbarh(height = 0, aes(xmin = correlation, xmax = 0)) + 
  geom_point() +
  labs(x = "How much this country's voting is correlated with the US",
       y = "")

country_correlations %>%
  group_by(country = item1) %>%
  summarize(median_correlation = median(correlation),
            avg_correlation = mean(correlation)) %>%
  arrange(desc(median_correlation)) %>%
  View()

country_correlations %>%
  filter(!is.na(continent1), !is.na(continent2)) %>%
  group_by(continent1, continent2) %>%
  summarize(avg_correlation = mean(correlation)) %>%
  arrange(desc(avg_correlation))

country_correlations %>%
  filter(!is.na(continent1), !is.na(continent2)) %>%
  filter(continent1 == continent2) %>%
  group_by(item1) %>%
  summarize(avg_intracontinent_correlation = mean(correlation)) %>%
  arrange((avg_intracontinent_correlation))

country_correlations %>%
  filter("German Democratic Republic" == item1)
```

```{r}
tt$issues %>%
  count(issue)

library(tidytext)

rc_words <- tt$roll_calls %>%
  filter(!is.na(short)) %>%
  unnest_tokens(word, short) %>%
  anti_join(stop_words, by = "word") %>%
  distinct(rcid, word) %>%
  add_count(word, name = "word_count") %>%
  filter(word_count >= 100)

unvotes %>%
  inner_join(rc_words, by = "rcid") %>%
  filter(country %in% c("Israel", "United States")) %>%
  group_by(word, country) %>%
  summarize_votes(min_votes = 100) %>%
  mutate(word = fct_reorder(word, pct_yes)) %>%
  ggplot(aes(pct_yes, word)) +
  geom_point(aes(size = n_votes, color = country)) +
  expand_limits(x = 0) +
  scale_x_continuous(labels = percent) +
  labs(x = "% yes",
       y = "")
```

```{r}
by_country_word <- unvotes %>%
  inner_join(rc_words, by = "rcid") %>%
  group_by(word, country) %>%
  summarize_votes(min_votes = 0)

library(tidytext)

by_country_word %>%
  widely_svd(word, country, pct_yes) %>%
  filter(dimension == 2) %>%
  mutate(word = reorder_within(word, value, dimension)) %>%
  top_n(30, abs(value)) %>%
  ggplot(aes(value, word)) +
  geom_col() +
  scale_y_reordered()
```

Recommend: topic modelling with the stm package (tidytext has a vignette)