Asymptotic Properties of Logrank and Stratified Logrank Tests on Highly Stratified Time-to-Event Data

• Activity Number: 345
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yabing Mai*+ and Sherry Liu and Tinghui Yu
• Companies: Merck Research Laboratories and Merck Research Laboratories and Binghamton University
• Keywords: Asymptotic property ; highly stratified data

Abstract

In clinical studies and statistical simulations, the loss of efficiency of logrank and stratified logrank tests was observed when the analysis of time-to-event data was highly stratified. Based on a comprehensive literature review, asymptotic properties of logrank and stratified logrank tests on highly stratified data does not appear to be adequately researched. In this joint work, we studied the asymptotic distribution and relative efficiency of logrank and stratified logrank tests on various types of highly stratified time-to-event data, especially when the number of strata increases as a function of the total sample size. In addition, we also explored the asymptotic property of the stratified logrank test in the more general case of the censoring distribution being dependent on both the treatment and covariates. Both analytical and simulation results will be provided in this work.