Clarifying xformla and weightsname

[raos-projects]

Hi Drs. Callaway & Sant'Anna,

Thanks for creating a great package. It's easy to use and so far I have no complaints.

I have some basic questions not evident from the documentation (to me) which if answered would settle my nerves (and hopefully others' too):

1. First, can you confirm that xformla and weightsname can be specified together in a call to the att_gt function? I have read Sant'Anna and Zhao (2020) Doubly Robust Difference-in-Difference estimators, and expect that these two arguments can be used together for doubly robust estimation. However, I haven't found an example online or in the vignettes which does this (in fact, I can't find any example which uses the weightsname argument.

2. Second, can you confirm what the formula in xformla is aiming to predict? Again from Sant'Anna and Zhao (2020), I expect it to be a predictor of average group-level outcomes for treatment group 'd' at time 't' (reading the explanation between equations 2.2 and 2.3, page 104). However, when I include an indicator variable for the treatment group (same variable specified in the gname argument) or an interaction term between time 't' and treatment group, I get a message that there are "No pre-treatment periods to test". I suspect this is due to collinearity of the interaction term within group-time cohorts and am happy to remove it, but it makes me question what I am predicting (am I predicting only the pre-treatment trends if not including a post-treatment indicator?) Some sources (slide 22 on this online PDF) claim xformla is used as the basis for the propensity score, but that feels wrong given your papers and the fact there is a weightsname argument.

3. Third, can you confirm that the weights specified in weightsname should be ATT weights in accordance with the DiD setup? And in theory, can ATE weights be used to recover an estimate of ATE? This is not expressed in any DiD papers to my knowledge but thought I'd ask.

Again, appreciate your time making such a useful tool!