#Load useful packages
library(tidyverse) #includes ggplot2, dplyr, readr, stringr
library(cowplot)
library(foreign)
library(gender)
library(knitr)
library(data.table)
library(reshape2)

#Read in data
data <- read.csv("https://raw.githubusercontent.com/drfreder/statscan-gender/master/statscan_11feb2018.csv")

#Subset to include only number, median, and average salaries of male and female faculty
data <- subset(data, RAN=="Male"|RAN=="Female")

#Subset and combine to re-format
data_num <- subset(data, STA=="Total teaching staff")
data_median <- subset(data, STA=="Median (dollars)")
data_average <- subset(data, STA=="Average (dollars)")
mdata <- cbind(data_num[,3:6],data_median[,3:6],data_average [,3:6])

#Make sure things match before deleting duplicated columns
mdata$Match <- mdata[ ,1] == mdata[ ,5]
mdata$Match2 <- mdata[ ,1] == mdata[ ,9]
tmp <- ifelse((all(mdata$Match) && all(mdata$Match2)), mdata <- mdata[,c(-5,-6, -9,-10, -13,-14)], FALSE)

#Rename columns
colnames(mdata) <- c("University", "Gender", "STA", "Total_staff", "STA.1", "Median", "STA.2", "Average")

#Remove unnecessary columns
mdata <- mdata[,c(-3,-5,-7)]

#Fix data types
mdata$Median <- as.numeric(as.character(mdata$Median))
mdata$Average <- as.numeric(as.character(mdata$Average))
mdata$Total_staff <- as.numeric(as.character(mdata$Total_staff))
mdata$University <- as.character(mdata$University)

#Fix accents in French university names
mdata$University <- enc2native(mdata$University)
mdata$University <- gsub(enc2utf8("<83>"),"E", mdata$University)
mdata$University <- gsub(enc2utf8("<8e>"),"e", mdata$University)
mdata$University <- gsub(enc2utf8("<88>"),"a", mdata$University)
mdata$University <- gsub(enc2utf8("<8f>"),"e", mdata$University)

#Omit rows that exclude medical and dental teaching staff
mdata <- mdata[!grepl(".*Excluding.*", mdata$University), ]
mdata$University <- gsub("- Including medical dental", "", mdata$University)

#Reshape data so that each university is a row
median_sum <- dcast(mdata, University~Gender, value.var="Median")
median_n <- dcast(mdata, University~Gender, value.var="Total_staff")
median_n$total <- median_n$Male+median_n$Female
median_sum$dif <- median_sum$Male - median_sum$Female
median_sum$percent <- median_sum$Female / median_sum$Male
median_sum$total <- median_n$total
median_n$percent_female <- median_n$Female / median_n$total
median_sum$percent_female <- median_n$percent_female

#Shorten one university name
median_sum[median_sum$University=="Universite du Quebec, Institut national de la recherche scientifique", "University"] <- "Universite du Quebec, INRS"

#Plot the gender wage gap, as a percent, for 66 Canadian institutions
p1 <- ggplot() +
      geom_point(data=median_sum, aes(x=reorder(University, -percent), y=as.numeric(percent*100)))+
      scale_y_continuous(limits=c(80, 100))+
      coord_flip()+
      xlab("Institution")+
      ylab("Gender pay gap (%)")+
      geom_segment(data=median_sum, aes(x=reorder(University, dif),
                   xend=reorder(University, dif), 
                   y=100, 
                   yend=as.numeric(percent*100)))+
      theme(axis.text.y = element_text(size=6))

#Plot the gender wage gap, in dollars, for 66 Canadian institutions
p2 <- ggplot(data=median_sum, aes(x=reorder(University, dif), y=as.numeric(dif)))+
      geom_point()+
      coord_flip()+
      xlab("Institution")+
      ylab("Gender pay gap ($)")+
      geom_segment(aes(x=reorder(University, dif),
                   xend=reorder(University, dif), 
                   y=0, 
                   yend=as.numeric(dif)))+
      theme(axis.text.y = element_text(size=6))

p1

p2

#Trim whitespace from university names
median_sum$University <- trimws(median_sum$University, which="both")

#Make a vector of U15 university names
U15 <- c("University of Alberta", "University of British Columbia", "University of Calgary", "Dalhousie University", "Universite Laval", "University of Manitoba","McGill University", "McMaster University", "Universite de Montreal", "University of Ottawa", "Queen's University", "University of Saskatchewan", "University of Toronto", "University of Waterloo","University of Western Ontario")

#Create a column indicating whether the university is in the U15
median_sum$U15 <- as.character(median_sum$University %in% U15)

#Subset to just the U15
U15 <- subset(median_sum, U15==TRUE)

#Calculate the mean gender pay gap for all non-U15 universities
nU15 <- subset(median_sum, U15==FALSE)
mean_dif_noU15 <- mean(nU15$dif)
se_dif_noU15 <- sd(U15$dif)/sqrt(length(nU15$dif))
U15red <- U15[c("University","dif")]
newRow <- data.frame(University="Non-U15 Universities", dif=mean_dif_noU15)

#Combine U15 data and means for non-U15 institutions
U15red <- rbind(U15red, newRow)

#Add French accents
U15red$University <- gsub("Universite Laval", "Universit\uE9 Laval", U15red$University)
U15red$University <- gsub("Universite de Montreal", "Universit\uE9 de Montr\u{E9}al", U15red$University)

#Plot the gender wage gap, in dollars, for the U15
p3 <- ggplot(data=U15red, aes(x=reorder(University, dif), y=as.numeric(dif)))+
      geom_point()+
      coord_flip()+
      xlab("Institution")+
      ylab("Gender pay gap ($/year)")+
      geom_segment(aes(x=reorder(University, dif),
                   xend=reorder(University, dif), 
                   y=0, 
                   yend=as.numeric(dif)))+
      theme(text = element_text(size=10),
            axis.text = element_text(size=10),
            axis.text.x = element_text(angle=90))
p3