B: Empirical Analysis

Here, we will extend the analyses we started in Homework $5$ to consider the questions of covariate balance and potential confounding. As we've stated, if Project STAR really was a clean RCT$,$ the answers to these questions should be clear. We will limit our analysis to the kindergarten class.

Overlap

Let's start by creating a table of summary statistics for the test score variable and key student and teacher characteristics. Again, only focusing on the kingergarten class.

First, create the test score variable and new female, white and freelunch indicators using the information in gender, ethnicity and lunchk. Note: Although ethnicity has six categories, all but a few students are either White or Black.

$```$

STAR STAR$2\%>\%$

mutate$($white $=$ case$\_$when$($ethnicity $==""$ ~ $1,$ TRUE ~ $0),$ #race

female $=$ case$\_$when$($gender $==""$ ~ $1,$ TRUE ~ $0),$ #gender

freelunch $=$ case$\_$when$($lunchk $==""$ ~ $1,$ TRUE ~ $0))$ #lunch

$```$

Next, use the st function to construct a table of summary statistics for the test$-$score variable and controls $($teacher experience experiencek, student gender $-$ female, race $-$ white, and eligibility for free lunch $-$ freelunch $)$ in the kindergarten classes.

$```$

st$($STAR$2,$ digits $=2,$ fixed.digits $=$ TRUE, numformat $=$ NA$,$

vars$=$c$(\text{'}\text{'},\text{'}\text{'},\text{'}\text{'},\text{'}\text{'},\text{'}\text{'}),$

group$=\text{'}\text{'},$ # review documentation$/$prior hw or slides for variable name

title$=$"Table $1."$

$)$

$```$