On Mon, 4 Apr 2016, Allin Cottrell wrote: One more point to mention in that context. For clustered standard errors in the fixed effects case we're now using the finite-sample adjustment described by Cameron and Miller (and used by Stata), namely sqrt((G/(G-1)) * (n-1)/(n-k)) where G is the number of groups or clusters, n is the total number of observations and k is the number of regressors. This is multiplied into the covariance matrix. It will make a relatively noticeable difference (larger std errors) if G is small. Allin