Many-to-one Confidence Intervals of Risk Ratios for Bilateral and Unilateral Correlated Data


Chang-Xing Ma, Shuyi Liang

Introduction


This article proposes simultaneous confidence intervals for many-to-one comparisons of risk ratios under Dalla’s model (1988). The construction of these confidence intervals relies on the profile likelihood, the score test, and the Wald-type test. The Bonferroni correction, the Holm–Bonferroni correction, and multivariate normal approximations are employed to adjust for the multiplicity when one makes several statistical inferences simultaneously. According to the simulation studies, the profile confidence interval with the Holm–Bonferroni correction is recommended as a result of robust coverage probabilities and satisfactory statistical powers. The application of the proposed methods are demonstrated via two real examples


Calculator


■ DATA


Please put your data in the box below,
or try the example 1 OR the example 2 in the paper.

5 × g   matrix separated by space or comma (g - number of groups)
3 × g   for Bilateral , 2 × g   for Unilateral
Note: the first group (1st column) is the reference group

■ OPTIONS


Significant level α: Default: 0.05, Confidence level = (1-α)*100
H0: θii0 (log risk ratios): g-1 vector of θi0, 0 corresponding no difference