Of the 6,325 kindergarten students who participated in the study, almost half or 3,052 were eligible for a free lunch program. The categorical variable sesk (1 == free lunch, 2 = non-free lunch) is a coarse proxy measure of family socio-economic status. If we think that children who begin school at some socio- economic disadvantage are likely to score lower on the standardized test of reading and arithmetic skills, then we should examine the proportion eligible for free lunch by type of class in kindergarten. Is the proportion elegible for a free lunch in the regular class with full-time aide statistically different from the proportion we observe in the other two types of class? If you think that there is a statistical difference, what effect would you predict the difference will have on the observed effect of class isze on test scores? Will the difference in the proportions of students eligible for free lunch amplify or attenuate the difference from class sizes?

Problem. We are not ready yet to address the main research question: does class size reduction improve kindergarteners' reading and arithmentic achieve- ments? We haven't dealth with regression involving categorical independent vari- ables. We can, however, consider the effect that teacher experience might have on their students' test score results.

I want you to study the connections between the kindergarten teacher's experi- ence, totexpk, and the end of year reading, treadssk, and arithmetic, tmathssk, test scores. Remember, a regression analysis entails more than issuing a Stata com- mand. Do carry out and exploratory analysis of each of the three variables, before you begin typing the command regress.