languages_values.R

library(lubridate)
library(rjson)
library(tidyverse)
library(ISOcodes)
library(ggplot2)
library(ggthemes)


setwd("~/Dropbox/ERC/templates/")

ess7 <- haven::read_dta("ESS7e02_1.dta")
data("ISO_639_2")
rownames(ISO_639_2) <- ISO_639_2$Alpha_3_B

ess7_values <- ess7 %>% 
  dplyr::select(idno, cntry, dweight, lnghom1, ipcrtiv:impfun) %>% 
  dplyr::mutate_at(-(1:4), ~ ifelse(.x > 6, NA, .x)) %>% 
  dplyr::mutate(language_home = ifelse(lnghom1 %in% c("777", "888", "999"), NA, lnghom1)) %>% 
  dplyr::mutate(language_home = ifelse(language_home == "GSW", "GER", language_home)) # Swiss German = German

ess7_values$language_home_iso2 <- ISO_639_2[tolower(ess7_values$language_home), 'Alpha_2']
ess7_values$language_home_name <- ISO_639_2[tolower(ess7_values$language_home), 'Name']

# Filter out small-time langauges
ess7_languages <- ess7_values %>% 
  dplyr::group_by(language_home_iso2) %>% 
  dplyr::summarise(n = n()) %>% 
  dplyr::filter(n > 200)
  
ess7_values <- ess7_values %>% 
  dplyr::filter(language_home_iso2 %in% ess7_languages$language_home_iso2)

# Show languages and counts in data
table(ess7_values$language_home_iso2, exclude = c())
# table(ess7_values$language_home_iso2, exclude = c()) %>% sort %>% barplot

summary(ess7_values)

ess7_values_means <- ess7_values %>% 
  dplyr::group_by(language_home_iso2) %>% 
  dplyr::summarise_if(~ is.numeric(.x), mean, na.rm = TRUE)


get_counts_perproject <- function(lang_iso2) {
  url <- "https://wikimedia.org/api/rest_v1/metrics/pageviews/aggregate/%s.wikipedia.org/all-access/all-agents/daily/20150701/20160701" %>%
    sprintf(lang_iso2)
  
  d <- rjson::fromJSON(file = url)$items %>% 
    dplyr::bind_rows() %>% 
    mutate(date = lubridate::as_date(timestamp, format = "%Y%m%d"))
  
  d  
  
}

page_names <- c("William_Shakespeare", "Game_of_Thrones")


get_counts_perpage <- function(lang_iso2, page_name) {
  
  url <- "https://wikimedia.org/api/rest_v1/metrics/pageviews/per-article/%s.wikipedia/all-access/all-agents/%s/daily/20150701/20160701" %>% 
    sprintf(lang_iso2, page_name)
  
  d <- rjson::fromJSON(file = url)$items %>% 
    dplyr::bind_rows() %>% 
    mutate(date = lubridate::as_date(timestamp, format = "%Y%m%d"))
  
  d  
}

pageviews_raw <- lapply(ess7_values_means$language_home_iso2, function(lang_iso2) {
  tryCatch(lapply(page_names, function(page_name) get_counts_perpage(lang_iso2, page_name)), 
           error = function(e) return(NULL))
})

pageviews_project <- lapply(ess7_values_means$language_home_iso2, function(lang_iso2) {
  tryCatch(get_counts_perproject(lang_iso2), 
           error = function(e) return(NULL))
})

d <- list()

for(iproject in seq_along(pageviews_raw)) {
  if(!is.null(pageviews_raw[[iproject]]))
    d[[iproject]] <- 
      dplyr::inner_join(pageviews_raw[[iproject]][[1]], pageviews_raw[[iproject]][[2]], by = c("date")) %>% 
      dplyr::inner_join(pageviews_project[[iproject]], 
                        by = c("date"))
}

d <- Reduce(bind_rows, d)

wiki_views <- d %>% 
  dplyr::mutate(views_raw_high = ifelse(is.na(views.x), 0, views.x),
                views_raw_low = ifelse(is.na(views.y), 0, views.y),
                views_project = views) %>%
  tidyr::separate(project, into = c("language", "project")) %>%
  select(language, date, starts_with("views_")) 

mean_interest <- wiki_views %>% 
  group_by(language) %>% 
  summarize(views_raw_high_avg = median(views_raw_high),
            views_raw_low_avg = median(views_raw_low),
            views_project_avg = median(views_project)) 


ess7_values_means$language <- ess7_values_means$language_home_iso2

interest_joined <- inner_join(mean_interest, ess7_values_means, by = "language")



interest_joined %>% 
  mutate(`Important to be creative` = cut(ipcrtiv, quantile(ipcrtiv, probs = c(0, 0.33, 0.5, 0.67, 1)))) %>% 
  ggplot(aes(views_raw_low_avg, views_raw_high_avg, label = language, 
             colour = `Important to be creative`, group = `Important to be creative`)) + 
  ggplot2::scale_x_log10() + ggplot2::scale_y_log10() + 
  ggplot2::xlab("Wikipedia pageviews: 'Game of Thrones'") + 
  ggplot2::ylab("Wikipedia pageviews: 'William Shakespeare'") +
  geom_smooth(method = "lm", se = F) +
  geom_text() + 
  scale_colour_brewer(palette = "RdYlBu") +
  ggplot2::ggtitle("Interest in Low versus high culture on Wikipedia\nis related to Schwarz human values") +
  theme_minimal(base_family = "Helvetica") + 
  theme(panel.border = element_rect(colour = "#00000055", fill=NA, size=0.25))


base_width <- 6.5
golden_ratio <- 1.7
ggsave("low_vs_high.pdf", width = base_width, height = base_width/golden_ratio)