Hi Stataliste'ers,

A more theoretical question here. I am currently working on cleaning my data for (PS)matching and it was my initial idea to match on characteristics from the control and treatment group. However, due to some missing data issues I will only have +/-7 variables to match on. This is of course not ideal with regards to bias, but luckily I do have access to some baseline characteristics of the treatment group as well. Would it be advised to add these characteristics in my current data set? I would than run a probit on the baseline characteristics of both the control and treatment and use the endline characteristics for psmatching.. I am just not sure if this is desirable and how I then have to merge the dataset as I will have two values (baseline and endline) for the treatment group.

I hope I have explained the problem clearly!

Linda