1_prepare_data.R

#!/usr/bin/env Rscript

library(tidyverse)

# Load individual files
child_has <- read_csv("data/raw_child_has_long.csv")
adult_has <- read_csv("data/raw_adult_has_long.csv")
child_sim <- read_csv("data/raw_child_sim_long.csv")
adult_sim <- read_csv("data/raw_adult_sim_long.csv")

# Just double checking that Document_name works well as a unique ID
# across data files
sanity_check <- function(d_has, d_similar) {
	stopifnot(nrow(d_has) == nrow(d_similar))
	stopifnot(length(unique(d_has$Document_name)) == length(unique(d_similar$Document_name)))
	demos_has <- d_has %>% select(Document_name, Age, Gender, Community_ID) %>% unique()
	demos_sim <- d_similar %>% select(Document_name, Age, Gender, Community_ID) %>% unique()
	demos_joined <- full_join(demos_has, demos_sim, by="Document_name")
	stopifnot(with(demos_joined, mean(Age.x == Age.y) == 1))
	stopifnot(with(demos_joined, mean(Gender.x == Gender.y) == 1))
	stopifnot(with(demos_joined, mean(Community_ID.x == Community_ID.y) == 1))
	return(TRUE)
}
sanity_check(child_has, child_sim)
sanity_check(adult_has, adult_sim)

# Combine responses for the two kinds of question into one tibble
combine_responses <- function(d_has, d_sim) {
	d_has <- d_has %>%
		mutate(type = str_sub(type_TF, 6, -1)) %>%
		rename(response_has = response_TF)
	d_sim <- d_sim %>%
		mutate(type = str_sub(type_similar, 9, -1))
	d_comb <- left_join(select(d_has, -type_TF),
		     select(d_sim, Document_name, type, response_similar),
		     by=c("Document_name", "type")) %>%
		rename(Participant_ID=Document_name)
}

child_comb <- combine_responses(child_has, child_sim)
adult_comb <- combine_responses(adult_has, adult_sim)

# Merge children and adults into one tibble, with a column for easy filtering
child_comb$AgeGroup = "children"
adult_comb$AgeGroup = "adults"
combined <- add_row(child_comb, adult_comb) %>%
	# Unique numeric PIDs
	mutate(Participant_ID = as.integer(as.factor(Participant_ID)),
	# Force NA for second question if answer to first is No
	       response_similar = if_else(!is.na(response_has) & response_has == "No", NA, response_similar),
	# Recode "no specification" gender as NA, after clarifying this is a non-response
	       Gender = if_else(Gender == "no specification", NA, Gender),
	# Rename "Galapagos_I" (not a country) to "Ecuador"
	       Country_full = if_else(Country_full == "Galapagos_I", "Ecuador", Country_full),
	       Community_ID = case_when(Community_ID == "Galapagos_urban" ~ "Ecuador_urban",
	                                Community_ID == "Galapagos_rural" ~ "Ecuador_rural",
					TRUE ~ Community_ID)) %>%
	# Just shuffling column order around to look "nice"
	select(Participant_ID, AgeGroup, Age, Gender,
	       Urb_rural, Country_full, Community_ID,
	       type, response_has, response_similar)
# Save this
write_csv(combined, "data/combined_long_data.csv")