psmatch2 postestimation regression using _weight: pweight vs. fweight

[Kai Blume]

Dear Statalist users,

I would like to run a fixed effects regression using a matched sample. I would like to use PSM with 3 nearest neighbors. My code looks as follows:

        psmatch2 treat match_var1 match_var2 match_var3  if year == 2019, neighbor(3) 
        xfill _weight, i(ID)
        xtreg satisfaction_total posttreatment##treat i.year [pweight=_weight], fe vce(cluster ID)

This seems to work fine. However, I have found different approaches how to deal with the weights provided by psmatch2. This source, e.g., uses

[fweight=_weight]

instead of pweights, but I also found forum entries suggesting pweights (and even aweights). Can anybody tell me whether pweights are appropriate? My problem is that fweights cannot be used after psmatch2 with 3 nearest neighbors (instead of one) as Stata doesn't allow noninteger frequency weights.

Thanks a lot!