Testing the Homogeneity of Two Survival Functions Against a Mixture Alternative Based on Censored Data |
| |
Authors: | Liting Zhu Xianming Tan |
| |
Institution: | NCIC Clinical Trials Group , Queen's University , Kingston, Ontario, Canada |
| |
Abstract: | When the survival distribution in a treatment group is a mixture of two distributions of the same family, traditional parametric methods that ignore the existence of mixture components or the nonparametric methods may not be very powerful. We develop a modified likelihood ratio test (MLRT) for testing homogeneity in a two sample problem with censored data and compare the actual type I error and power of the MLRT with that nonparametric log-rank test and parametric test through Monte-Carlo simulations. The proposed test is also applied to analyze data from a clinical trial on early breast cancer. |
| |
Keywords: | Censored data Hypothesis testing Mixture model Survival functions |
|
|