A π΄ race: The Wild Cluster Bootstrap vs Satterthwaite-corrected Sandwich Estimators when the number of Clusters is small
A couple of months ago, Gordon Burtch shared an excellent Twitter thread on the merits of wild cluster bootstrap inference when the regression error terms are clustered into a small group of clusters:
How biased are clustered SEs with 'few' clusters? A simulation illustrating this. DGP is y~x, 50 clusters, ...
[Read more...]