Question about adding group-specific linear trend in regression

[Austin Yi]

Hello Stata community,

I would like to ask a practical question about how to add group-specific linear trend in regression. Normally we would just add c.year#i.group to allow different groups having different linear trends. However, this would fit the linear trends using post-treatment data points. Is there a way to fit the linear trends with only pre-treatment data points (and extrapolating to the post treatment period)? Thank you.

[Fabian Gruenwald]

Hello Austin,
did you find a solution since then?
I am currently looking for a way to do this aswell.

[Xiao Zhang]

one solution might be to estimate the first stage regression using the pre- treatment sample, then get the residual for both pre- and post- treatment period. lastly, run the second stage regression with the full sample, using the residual from the last step as the DV. Not sure if this is econometrically correct, but intuitively you will get what you need.