Asymptotic Multivariate Confidence Rectangles and Multiple Comparisons Methods for Difference of Vectors of Proportions

Asymptotic confidence rectangles for the difference of 2 vectors of proportions are proposed and their asymptotic probability of the coverage is proved. Their performance is compared by means of simulations with the performance of the asymptotic confidence regions constructed by the Bonferroni principle from the Agresti-Caffo confidence intervals. The performance of the resulting multiple comparison methods for detecting coordinates of vectors of proportions having different values is also illustrated by simulations.