Syndicate content

Add new comment

Submitted by David Reinstein on

Thanks for the post. To be clear, are you suggesting that "demean[ing] (or center) your covariates and fully interact them with your treatment indicator" itself will yield a coefficient representing the ATE? Is there a simple proof of that somewhere? If that is all that needs doing, what is the need for the more involved procedure (and R/Stata code) that Gibbons et al provide? Sorry if I'm being dense here.