Sample Size Calculations for Testing Equivalence of Two Exponential Distributions With Right Censoring: Allocation With Costs

Abstract

The present study considered two independent exponential distributions with the hypothesis for testing equivalence of lifetime means or failure rates, and aimed to determine the required uncensored sample size based on power, sampling cost, and censoring proportion simultaneously in the case of right censoring. Approximate sample size formulas with an iterative procedure were proposed and an uncensored sample size allocation ratio was derived to minimize the total cost for a designated power or maximize statistical power for a limit cost. R codes are provided for easy application. The proposed methods are validated in terms of Type I errors and statistical power in a simulation study, and are recommended for the future use.