Fit indicators for the sdid command

[Sam Fairless]

Dear all,

I am currently writing a thesis on health insurance in the USA, I have used the sdid command with the placebo option. This returns the estimates for: ATT, Std. Error, t-value, p-value, and the 95% confidence interval.

I added this to my thesis and showed it to my supervisor, who requested I add an adjusted r-squared value to indicate fit. The documentation of the command does not show an r-squared estimate, which makes sense. I thought of adding a RMSE value, but am at a loss on how that would fit in the code. Does anyone have any experience on producing fit indicators with the sdid command?

Many thanks,
Samuel Fairless

[Frank Huang]

I believe that SDID does not require reporting an R-squared at all. However, if one insists on constructing an R-squared value, a possible approach would be to save the unit and time weights from the sdid command and then replicate the weighted two-way fixed effects regression described in the SDID paper.