Regression discontinuity design (RDD) using STATA

[Man Yang]

Hello all, I'm learning to use RDD on a current project I'm working on. Basically, I want to see whether the poverty label impact students' future possibility of dropping out. Although it seems there are specific command such as -rd- that can be used to apply RDD, I was told that RDD could be carried out using just regular regression model, which in my case it would be logistic model I suppose because the outcome I am interested in is a binary variable (1=drop out, 0= didn't drop out). In my data, there is a continuous variable "poverty" which reflects some sort of SES index for instance and the cut-score I am using is 540. That is any student with poverty variable less than 540 will be labeled as in poverty and those above 540 will be labeled not in poverty. Following the example in this link: https://stats.idre.ucla.edu/stata/ex...tter/chapter9/ , can I set up my RDD model as the follows?

gen newvar = poverty - 540
gen inpover = 0
replace inpover = 1 if poverty >=540
logit dropout newvar inpover

Thanks,
Man

[River Huang]

RDD with discrete outcomes may need additional treatment. Please have a look of the following paper http://www.sciencedirect.com/science...04407617301471. The website of the author provides R/Matlab code for applying the method.