A Generalization of a Nonparametric Test for Stochastically Ordered Distributions to Censored Survival Data

Journal of the Royal Statistical Society: Series B (Statistical Methodology) · 1988 · Volume 50 · Issue 3

DOI: 10.1111/j.2517-6161.1988.tb01737.x

Abstract

Currently used test statistics for stochastically increasing survival times in ak-sample problem are linear combinations of either Gehan (generalized Wilcoxon) or generalized Savage score sums. The power of these tests depends ona prioriknowledge about the true form of the alternative. When isotonic regression is applied to these score sums, it yields a new trend test for right-censored survival data which generalizes to left-truncated survival data and to counting processes. In a Monte Carlo study, its empirical power is at least comparable with that of the linear trend tests in their standard form (without prior information), and it is markedly superior if only the last survival distribution turns out to be stochastically smaller or larger than the control distribution.

Keywords

amalgamation, counting processes, left truncation, stochastic order, trend tests

Enable JavaScript for citation tools and additional paper interactions.