Abstract
Unilateral or bilateral paired data are often encountered in medical studies. It is natural to analyze such data by combining them. In the article, the homogeneity test of risk difference and sample size determination are investigated based on the stratified uni- and bilateral paired data. Under Dallal’s model, we propose six statistical tests and eight sample size methods. Monte Carlo simulations show that the score test behaves well in various parameter settings. Wald-type test always has higher empirical powers. Iterative algorithm and asymptotical method are introduced to estimate sample sizes, and the estimated powers of the former are closer to the desired power than that of the latter. An ophthalmology example is used to demonstrate the performance of the proposed methods.