Our method, for the special case of comparing groups,
is particularly simple,
gives asymptotically correct results as the number of clusters in each group tends to infinity,
and can be implemented using any standard computer program for the analysis of independent binary data after a small amount of preprocessing.