03_bivariate_analysis.Rmd

---
title: "03_bivariate_analysis"
author: "Anja Rainer"
date: "16 7 2021"
output: html_document
---

```{r setup, include=FALSE}
library(tidyverse)
library(RColorBrewer)

knitr::opts_chunk$set(echo = TRUE, warning = FALSE, dpi = 300)

df <- tar_read(with_countries)

question_codes <- read_csv("data/processed/question_codes.csv")
```

# Merge dataframes

```{r}

# bind_cols()
# bind_rows()


```

# Bivariate analysis Gender

```{r}
# omit "this topic is not relevant" and "don't know" -> these will be turned
# to NA
answer_levels <- c("Strongly disagree", "Disagree", "Neither agree nor disagree",
                   "Agree", "Strongly agree")

gender_df <- df %>% 
  select(starts_with("A10["), E2) %>% 
  mutate(across(starts_with("A10["), .fns = factor, levels = answer_levels)) %>% 
  # remove "other" genders;
  # potentially could also remove "prefer not to say", since cell counts are low
  filter(E2 != "Other")

#gender_df %>%
 # pivot_longer(starts_with("A10[")) %>% 
  # count(E2, name, value) %>% 
  # filter(!is.na(value)) %>% 
  # group_by(E2, name) %>% 
  # mutate(prop = n / sum(n)) %>% 
  # ggplot(aes(E2, prop, fill = value,)) +
  # geom_col() +
  # coord_flip() +
  # scale_fill_manual(values = c("#FFDB6D", "#C4961A", "#F4EDCA", 
                "#D16103", "#C3D7A4", "#52854C", "#4E84C4", "#293352")) +
  # facet_wrap(vars(name))

gender_df %>%
  pivot_longer(starts_with("A10[")) %>% 
  count(E2, name, value) %>% 
  filter(!is.na(value)) %>% 
  group_by(E2, name) %>% 
  mutate(prop = n / sum(n)) %>% 
  ggplot(aes(E2, prop, fill = value,)) +
  geom_col() +
  coord_flip() +
  scale_fill_brewer(palette = "PuOr") +
  scale_y_continuous(labels = function(x) scales::percent(x, accurarcy = 1)) +
  facet_wrap(vars(name)) +
  guides(fill = guide_legend(title = NULL, nrow = NULL))


```

# A1 OS practices by country

```{r, fig.width=8, fig.height=4}
# factor out function to reduce repetition
prepare_per_country <- function(df, vars, by, fac_levels) {
  df %>% 
    select({{vars}}, {{by}}) %>% 
    mutate(across({{vars}}, .fns = factor, levels = fac_levels)) %>% 
    mutate(E1_rec = forcats::fct_lump_min({{by}}, min = 10, 
                                          other_level = "Other")) %>%
    filter(E1_rec != "NA")
}



country_a1_df <- prepare_per_country(df, starts_with("A1["), E1_rec,
                                     answer_levels)

country_a1_df %>%
  pivot_longer(starts_with("A1[")) %>% 
  count(E1_rec, name, value) %>% 
  filter(!is.na(value)) %>% 
  group_by(E1_rec, name) %>% 
  mutate(prop = n / sum(n)) %>% 
  ggplot(aes(E1_rec, prop, fill = value,)) +
  geom_col() +
  coord_flip() +
  scale_fill_brewer(palette = "PuOr") +
  facet_wrap(vars(name)) +
  scale_y_continuous(labels = function(x) scales::percent(x, accurarcy = 1)) +
  guides(fill = guide_legend(title = NULL, nrow = NULL)) +
  labs(x = NULL, y = NULL, title = NULL)


```

# B5 hours of training by country

```{r,fig.width=8, fig.height=4}

answer_levels_b5 <- c("None", "1", "2", "3-5", "More than 5")

country_b5_df <- prepare_per_country(df, B5, E1_rec, answer_levels_b5)
  

country_b5_df %>%
  pivot_longer(B5) %>% 
  count(E1_rec, name, value) %>% 
  filter(!is.na(value)) %>% 
  group_by(E1_rec, name) %>% 
  mutate(prop = n / sum(n)) %>% 
  ggplot(aes(E1_rec, prop, fill = value,)) +
  geom_col() +
  coord_flip() +
  scale_fill_brewer(palette = "PuOr") +
  facet_wrap(vars(name)) +
  scale_y_continuous(labels = function(x) scales::percent(x, accurarcy = 1)) +
  guides(fill = guide_legend(title = NULL, nrow = NULL)) +
  labs(x = NULL, y = NULL, title = NULL)

```

# B10 Preferred way to learn OS by country

```{r, fig.width=8, fig.height=4}

answer_levels_3 <- c("Yes", "No")

country_b10_df <- prepare_per_country(df, starts_with("B10["), E1_rec,
                                      answer_levels_3)

country_b10_df %>%
  pivot_longer(starts_with("B10[")) %>% 
  count(E1_rec, name, value) %>% 
  filter(!is.na(value)) %>% 
  group_by(E1_rec, name) %>% 
  mutate(prop = n / sum(n)) %>% 
  ggplot(aes(E1_rec, prop, fill = value,)) +
  geom_col() +
  coord_flip() +
  scale_fill_brewer(palette = "Set2") +
  facet_wrap(vars(name)) +
  scale_y_continuous(labels = function(x) scales::percent(x, accurarcy = 1)) +
  guides(fill = guide_legend(title = NULL, nrow = NULL)) +
  labs(x = NULL, y = NULL, title = NULL)


```

# C8 Degree of support in practicing OS at institution by country

```{r, fig.width=8, fig.height=4}

answer_levels_7 <- c("Don’t know/ Don’t have enough information", "I do not receive any support or incentives", "I do not receive any support or incentives but would like to", "I receive some support or incentives", 
"I receive sufficient support or incentives")

country_c8_df <- prepare_per_country(df, starts_with("C8["), E1_rec,
                                      answer_levels_7) %>%
              left_join(question_codes, by = c(starts_with("C8[") = "short_code")) %>% 


country_c8_df %>%
  pivot_longer(starts_with("C8[")) %>% 
  count(E1_rec, name, value) %>% 
  filter(!is.na(value)) %>% 
  group_by(E1_rec, name) %>% 
  mutate(prop = n / sum(n)) %>% 
  ggplot(aes(stringr::str_wrap(question_specification, 40), E1_rec, prop, fill = value,)) +
  geom_col() +
  coord_flip() +
  scale_fill_brewer(palette = "PuOr") +
  facet_wrap(vars(name)) +
  scale_y_continuous(labels = function(x) scales::percent(x, accurarcy = 1)) +
  guides(fill = guide_legend(title = NULL, nrow = NULL)) +
  labs(x = NULL, y = NULL, title = NULL)

```