Jhoanne Marsh Gatpatan, PhD Candidate
Department of Statistics
Western Michigan University
Abstract:
Martinez and Naranjo (2010) proposed a diagnostic Q
that may be used as a pretest for deciding between the
logrank or Wilcoxon test in the two-sample case. However,
we often want to compare treatment groups
after adjusting for covariate effects. In this study,
we investigate the performance of the stratified logrank and
stratified Wilcoxon tests in controlling for the effects of
a categorical covariate.
We propose pretests based on Q that help
decide between using stratified logrank or
stratified Wilcoxon. Power perfomance of each adaptive
test was assessed using simulations under various patterns of
proportional hazard and non-proportional hazard distributions.
---
All statistics graduate students are expected to attend.