## 2. ESTIMATING CAUSAL EFFECTS WITH RANDOMIZED EXPERIMENTS

## 2.5 DO SMALL CLASSES IMPROVE STUDENT PERFORMANCE?

## Set the working directory
setwd("~/Desktop/DSS") # example of setwd() for Mac 
setwd("C:/user/Desktop/DSS") # example for Windows

## Load the dataset
star <- read.csv("STAR.csv") # reads and stores data

## Look at the data
head(star) # shows first observations

## Relational operators in R
3==3 # example of a TRUE statement
3==4 # example of a FALSE statement
star$classtype=="small" 

## Creating new variables
ifelse(star$classtype=="small", 1, 0) # creates contents of variable
star$small <- ifelse(star$classtype=="small", 1, 0) # creates new variable
head(star) # shows first observations

## Subsetting variables
mean(star$reading) # calculates the mean of reading
mean(star$reading[star$small==1]) # mean of reading for treatment group
mean(star$reading[star$small==0]) # mean of reading for control group

## Compute the difference−in−means estimator for reading 
mean(star$reading[star$small==1]) - mean(star$reading[star$small==0])

## Compute the difference−in−means estimator for math 
mean(star$math[star$small==1]) - mean(star$math[star$small==0])

## Compute the difference-in-means estimator for graduated
mean(star$graduated[star$small==1]) - mean(star$graduated[star$small==0])

## To help with interpretation
mean(star$graduated[star$small==1]) # mean of graduated for treatment group
mean(star$graduated[star$small==0]) # mean of graduated for control group
0.8735043 - 0.8664731 # difference-in-means estimator for graduated

## Uncertainty
hist(star$reading[star$small==1])
hist(star$reading[star$small==0])

boxplot(star$reading~star$small)

test <- t.test(star$reading~star$small)
test

## Tidyverse approach
library(tidyverse)

star %>%
  mutate(small2 = if_else(classtype=="small",1,0)) %>%
  group_by(small2) %>%
  summarise(reading_avg = mean(reading),
            math_avg = mean(math),
            grad_avg = mean(graduated)) %>%
  mutate(reading_diff=diff(reading_avg),
         math_diff=diff(math_avg),
         grad_diff=diff(grad_avg))

star %>%
  ggplot(aes(y=reading, x=classtype))+
  geom_boxplot()

library(ggstatsplot)

star %>%
  ggbetweenstats(small, reading)

star %>%
  ggbetweenstats(small, reading,
                 type="robust")