Online Program Home
My Program

Abstract Details

Activity Number: 258 - SPEED: Causal Inference and Related Methodology
Type: Contributed
Date/Time: Monday, July 30, 2018 : 2:00 PM to 2:45 PM
Sponsor: Section on Statistics in Epidemiology
Abstract #333020
Title: Maximum Likelihood Estimation of the K Parameter in the Poly-K Trend Test for Time-to-Event Data
Author(s): Anna Korpak* and Barbara McKnight
Companies: VA ERIC and University of Washington
Keywords: carcinogenicity; poly-k; poly-3; cancer; MLE; Weibull
Abstract:

In many animal carcinogenicity experiments, tumors are detected post-mortem and higher doses may be toxic. When toxicity induces mortality differences by treatment group, many statistical methods for survival data perform poorly. The poly-k test, developed by Bailer & Portier (1988), avoids problems from differential mortality and is commonly used. The poly-k test performs best when there is a Weibull tumor incidence hazard with shape parameter k (usually k=3); it can be biased under treatment toxicity when the assumed k is wrong. One solution is to estimate k. Moon et al. (2003) derived a method based on estimating lifetime cumulative tumor incidence rates; to perform well, it requires experiments with multiple interim sacrifices. We introduce an alternative estimator of k, obtained by maximizing the full data likelihood. Under simulations with varying tumor hazards and sacrifice designs, the resulting poly-k(MLE) test has similar type I error and power to the poly-k test with correctly-specified k; it maintains size better, with comparable or higher power, than the test using the Moon et al. estimate. The MLE-based test performs well in settings with no interim sacrifices.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program