Evaluating the Performance of Simultaneous Stepwise Confidence Intervals for the Difference Between Two Poisson Rates

• Activity Number: 167
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Brianna Bright*+ and Julia Soulakova
• Companies: University of Nebraska-Lincoln and University of Nebraska-Lincoln
• Keywords: multiple testing ; simultaneous confidence intervals ; Poisson distribution

Abstract

We consider the problem of simultaneously estimating Poisson rate differences via applications of the Hsu and Berger stepwise confidence interval method, where comparisons to a common reference group are performed. While the original method was presented in the context of determining the minimum effective dose in drug trials in terms of normally distributed data, the procedure is well suited for any model where the parameters monotonically increase or decrease over some specified groups. Applications of this method to Poisson-distributed outcomes have not been addressed in the literature. We discuss continuity-corrected confidence intervals around Poisson rate differences and investigate the Hsu and Berger method performance in conjunction with a moment-based confidence interval, and uncorrected and corrected for continuity Wald and Pooled confidence intervals. Using simulations, we compare nine individual confidence intervals in terms of coverage probability and the Hsu and Berger method with the nine confidence intervals in terms of family-wise error rate and overall and local power. The simulations show that these statistical properties depend highly on parameter settings.