Syndicate content

Add new comment

Submitted by Rahul Mehrotra on

Hi David, thanks for this useful blog-post. One overall question which I don't think you have discussed before (apologies if I missed it): what do you think about how randomization inference should be operationalized, so to speak, in the randomized experiments literature? Should we start reporting two sets of p-values in results tables, as we did with robust and cluster(-bootstrapped) after that approach gained prominence a decade ago? And if there are significant differences then interpret result as not robust?