Effect of unequal censoring on the size and power of the logrank and Wilcoxon types of tests for survival data.

摘要: This paper examines the effect of random unequal censoring on size and power two-sample logrank Wilcoxon types tests for comparing two survival distributions by simulation with small samples from censored exponential distributions. We compared equal-sized n = 8, 16, 32 1000 (size) 500 (power) trials 16 combinations proportions 0, 20, 40, 60 per cent in each samples. For asymptotic normality (AN), Peto-Peto, Wilcoxon-type performed at nominal 5 expectations, but Mantel test exceeded acceptance region 6 combinations. 32, all showed proper size, Peto-Peto being most conservative presence censoring. powers hazard ratios 1.4 2.0. The Mantal 90 to 95 efficiency relative parametric AN test. Both identically had lowest examined appeared robust differential patterns studied. A modified version comparable