Ben Porter posted on Monday, November 21, 2016 - 3:45 pm
I ran a GMM on a large dataset with four time points. A reviewer has requested that we propensity adjust or match upon propensity of response. My understanding was that non-response would be taken into account through the ML estimation and each observation's propensity score would be equal to 1. Is my thinking incorrect about this matter?
I assume that you have a treatment variable and that you have a non-experimental study. The propensity score is the probability of being in the treatment group as a function of covariates that make the treatment and control groups different. Regressing on the propensity score allows a meaningful comparison on the outcome of the two groups.