analysis.Rmd

---
title: "analysis"
author: "Benedikt Ströbl"
date: "2022-11-29"
output: html_document
---
```{r}
require(tidyverse)
require(gtools)
require(gtable)
require(gt)
require(gtExtras)
require(gtsummary)
require(ggforce)
require(stargazer)
require(modelsummary)
require(showtext)
font_add_google("Roboto Condensed", "Roboto Condensed")
showtext_auto()
source("ggplot_themes.R")

speeches <- read_delim("data/speeches_final.csv", 
    delim = ";", escape_double = FALSE, trim_ws = TRUE, show_col_types = F) %>% 
  mutate(
    lr_cat_5 = quantcut(lrgen),
    lr_cat_5 = case_when(
      lr_cat_5 == "[3.21,4.46]" ~ "left",
      lr_cat_5 == "(4.46,5.86]" ~ "center_left",
      lr_cat_5 == "(5.86,6.91]" ~ "center_right",
      lr_cat_5 == "(6.91,8.33]" ~ "right",
      TRUE ~ "independ."
    ),
    lr_cat_2 = case_when(
      lrgen < 3 ~ "left",
      lrgen >= 7 ~ "right",
      TRUE ~ "Other"
    )
  )

speeches %>% 
  glimpse()
```

## General Time Series

### Count Speeches over the years

```{r}
speeches %>% 
  group_by(
    date = lubridate::floor_date(date, 'bimonthly')
  ) %>% 
  summarise(
    count = n()
  ) %>% 
 ggplot(aes(
    y = count,
    x = date
  )) + 
  geom_line(color = "darkblue", size = .7) + 
  labs(
    title = "Number of Speeches Held by the Current EU College Members",
    subtitle = "Count of recorded public appearance be two-month period (bimonthly)",
    caption = "Source: Own Illustration",
    x = "Time",
    y = "Nr. of Speeches"
  ) +
  theme_custom()
```

### Count of speeches during 2020



```{r}
speeches %>% 
  filter(year == 2020) %>% 
  group_by(
    date = lubridate::floor_date(date, 'weekly')
  ) %>% 
  summarise(
    count = n()
  ) %>% 
 ggplot(aes(
    y = count,
    x = date
  )) + 
  geom_line(color = "darkblue", size = .7) + 
  labs(
    title = "Number of Speeches Held by the Current EU College Members",
    subtitle = "Count of recorded public appearance be two-month period (bimonthly)",
    caption = "Source: Own Illustration",
    x = "Time",
    y = "Nr. of Speeches"
  ) +
  theme_custom()
```

### Cognitive vs. Affective Language over the years

```{r}
speeches %>% 
  group_by(
    date = lubridate::floor_date(date, 'month')
  ) %>% 
  
  select(date, 
         emotion_score, 
         emotion_score_own_embedding, 
         emotion_score_own_embedding_tfidf,
         emotion_score_tfidf,
         affect_score,
         affect_score_own_embedding,
         affect_score_own_embedding_tfidf,
         affect_score_tfidf,
         cognition_score,
         cognition_score_own_embedding,
         cognition_score_own_embedding_tfidf,
         cognition_score_tfidf)  %>% 
  summarize_all(
    mean
  ) %>%  
  ungroup() %>% 
  pivot_longer(-date, names_to = "measure", values_to = "value") %>% 
  filter(measure %in% c('cognition_score_own_embedding_tfidf', 'affect_score_own_embedding_tfidf')) %>% 
  mutate(measure = if_else(measure == 'cognition_score_own_embedding_tfidf', 'Cognition Score', 'Affect Score')) %>%
  mutate(
    president = if_else(date < "2019-12-01", "Juncker", "von der Leyen")
  ) %>% 
  ggplot(aes(
    x = date,
    y = value,
    color = measure,
    shape = president
  )) +
  geom_line() +
  labs(
    title="Cognitive/Affective Language by the EU Commission",
    subtitle = "How did the current college members of the EU Commission communicate?",
    caption = "Source: Own Illustration",
    x = "Time",
    y = "Score",
    color = "Measure"
  ) + 
  annotate("rect", fill = "grey", alpha = 0.5, 
        xmin = as.Date("2019-12-01"), xmax = as.Date("2022-11-26"),
        ymin = -Inf, ymax = Inf) +
  annotate("text", x = as.Date("2019-12-01"), y = 0.65, label = "v. d. Leyen", size = 6) +
  scale_color_brewer(palette = "Dark2") +
  geom_smooth(method = "lm") +
  theme_custom()
```

```{r}
over_time <- speeches %>% 
  group_by(
    date = lubridate::floor_date(date, 'month')
  ) %>% 
  
  select(date, 
         emotion_score, 
         emotion_score_own_embedding, 
         emotion_score_own_embedding_tfidf,
         emotion_score_tfidf,
         affect_score,
         affect_score_own_embedding,
         affect_score_own_embedding_tfidf,
         affect_score_tfidf,
         cognition_score,
         cognition_score_own_embedding,
         cognition_score_own_embedding_tfidf,
         cognition_score_tfidf)  %>% 
  summarize_all(
    mean
  ) %>%  
  ungroup() %>% 
  pivot_longer(-date, names_to = "measure", values_to = "value") %>% 
  filter(measure %in% c('cognition_score_own_embedding_tfidf', 'affect_score_own_embedding_tfidf')) %>% 
  mutate(measure = if_else(measure == 'cognition_score_own_embedding_tfidf', 'Cognition Score', 'Affect Score')) %>% 
  mutate(
    president = if_else(date < "2019-12-01", "Juncker", "von der Leyen")
  )


qplot(
  x = date,
  y = value,
  color = measure,
  shape = president,
  data = over_time,
  geom=c("point", "smooth"))
```



### Focus: Juli / August 2021: Afghanistan Coup

```{r}
speeches %>% 
  filter(
    year == 2021
  ) %>% 
  group_by(
    date = lubridate::floor_date(date, 'month')
  ) %>% 
  
  select(date, 
         emotion_score, 
         emotion_score_own_embedding, 
         emotion_score_own_embedding_tfidf,
         emotion_score_tfidf,
         affect_score,
         affect_score_own_embedding,
         affect_score_own_embedding_tfidf,
         affect_score_tfidf,
         cognition_score,
         cognition_score_own_embedding,
         cognition_score_own_embedding_tfidf,
         cognition_score_tfidf)  %>% 
  summarize_all(
    mean
  ) %>%  
  ungroup() %>% 
  pivot_longer(-date, names_to = "measure", values_to = "value") %>% 
  filter(measure %in% c('cognition_score_own_embedding_tfidf', 'affect_score_own_embedding_tfidf')) %>% 
  mutate(measure = if_else(measure == 'cognition_score_own_embedding_tfidf', 'Cognition Score', 'Affect Score')) %>% 
  ggplot(aes(
    x = date,
    y = value,
    color = measure
  )) +
  geom_line(size = 0.7) +
  labs(
    title = "Affect and Cognition Scores in 2021",
    subtitle = "Lines are depicting the mean monthly scores over all speeches",
    caption = "Source: Own illustration",
    x = "Date",
    y = "Score",
    color = "Measure"
  ) +
  scale_color_brewer(palette = "Dark2") +
  theme_custom()
```

### Male Female distribution

```{r}
speeches %>% 
  ggplot(aes(
    y = cognition_score_tfidf,
    x = sex,
    color = sex
  )) + 
  geom_violin(aes(color = sex), size = 1, alpha = .5) +
  geom_jitter(alpha = .25, width = .3) +
  geom_sina(alpha = .25) +
    geom_boxplot(outlier.alpha = 0, coef = 0,
                 color = "gray40", width = .2) +
  scale_color_brewer(palette = "Dark2", guide = "none") +
  labs(
    x = "Gender",
    y = 'Score',
    title = "Cognitive/Rational Language Across Sex",
    subtitle = "Distribution of all speeches' cognition score held by male/female commissioners in data set",
    caption = "Data source: Own Illustration"
  ) +
  coord_flip() +
  theme_custom()

```


## Regression Model (Gender, Ideologoy, EU Position)

```{r}
sex <- lm(scale(emotion_score_own_embedding)~sex, data=speeches)

lr <- lm(scale(emotion_score_own_embedding)~scale(lrgen), data=speeches)

complete <- lm(scale(emotion_score_own_embedding)~ sex + scale(lrgen), data=speeches)


stargazer(sex, lr, complete, out = "regression.html",
          dep.var.caption = "Emotionality Score",
          dep.var.labels = "")
```


```{r}
speeches %>%
  ggplot(aes(
    x = sex,
    y = emotion_score_tfidf
  )) +
  geom_boxplot()
```