Linear Regression for pre-post test

[Polina Qwerty]

Hi everyone,

I would like some help for a linear regression model. I'm trying to compare the outcome (scores) between 2 academic years based on participation in a program.

Would the command be: by 'program', sort : regress 'post-score var' 'pre-score var' ? Or is it: regress 'post-score var' 'program' 'pre-score var' ?

I would really appreciate any help!

Thank you

[Fei Wang]

Polina, more details about your data structure would help. I guess you have a student-level data with three variables: pre-score, post-score, and an indicator of program participation (1 = participants, 0 = non-participants). You may run the following codes.

gen dscore = postscore - prescore
reg dscore program

The codes above are essentially a DiD: The effect of program participation on score = score change for participants - score change for non-participants. In other words, non-participants are treated as a reference group for participants. But if you'd like to simply compare scores between two years only for participants, just run "reg dscore if program == 1", and the constant term would be the answer -- but it may not be interpreted as the effect of program on scores as something beyond the program may affect scores at the same time, and that's why non-participants are necessary to be the reference group.

[Polina Qwerty]

Thanks so much Fei! This was really helpful.