Some of the earliest posts on this blog concerned the inferential challenges of cluster randomized trials when clusters are few in number (see here and here for two examples of discussion). Today’s post continues this theme with a focus on better practice in the treatment of standard errors.
