I am trying to estimate the difference in the effect of a policy on an outcome with many 0's between 2 groups. To control for differences in baseline characteristics between my two groups before the policy was implemented, I have constructed overlap weights (https://jamanetwork.com/journals/jam...stract/2765748) using pre-policy data for each individual in my dataset.
I would like to use the twopm command in Stata 17 and apply my overlap weights using [pweight=overlapweight], but I have a question about the validity of doing so. I see in the Stata Journal article describing the twopm command (https://journals.sagepub.com/doi/pdf...867X1501500102) that the command is compatible with pweights, but I am wondering about if the weights I have constructed are acheiving their intended purpose given the second part model is only using a subset of the observations (non-0's) and my weights were created using the entire sample.
Is anyone able to provide any insight on whether applying my weights in this way with the twopm is valid? Or if not if there are other ways to use these weights with the twopm command?
I would like to use the twopm command in Stata 17 and apply my overlap weights using [pweight=overlapweight], but I have a question about the validity of doing so. I see in the Stata Journal article describing the twopm command (https://journals.sagepub.com/doi/pdf...867X1501500102) that the command is compatible with pweights, but I am wondering about if the weights I have constructed are acheiving their intended purpose given the second part model is only using a subset of the observations (non-0's) and my weights were created using the entire sample.
Is anyone able to provide any insight on whether applying my weights in this way with the twopm is valid? Or if not if there are other ways to use these weights with the twopm command?
