Inference About the Intraclass Correlation Coefficient in the Beta-Binomial Anova for Proportions

Summary The paper is concerned with the analysis of sets of proportions where there is strong variation within sets. In one parameterization (Crowder, 1978) the underlying mean proportions are π1,…, πk in the k sets, and variation within sets is associated with a parameter σ2. The last of these is the intraclass correlation coefficient for the basic binary data and is of independent interest, rather than just a nuisance parameter. We propose a technique for inference about σ2.