The Empirical Analysis Of Non-parametric Test Statistics In Survival Analysis

There are many two-sample and non-parametric test statistics used to test the rightcensored data in survival analysis.However,the test results of different test statistics are different.Our aim in this paper is to compare various two-sample non parametric test statistics for right censored survival data.We generate survival data with different characteristics in the simulation.Specifically,four pairs of risk functions in scenario(1)are considered where the hazard functions are equal,proportional,and intersecting at one point or neither proportional nor intersecting.The four pairs of risk functions in scenario(2)are all designed to have cross property.While in scenario(3),the risk functions have obvious differences in the early,middle and late stages.We further consider the influence of sample size and censorship rate on the test results.After fully understanding the non-parametric test statistics in the numerical simulation,we apply them in real data analysis.The contents of the paper are as follows.Section 1 includes the background of the non-parametric tests,and briefly introduces the related theoretical knowledge involved in this paper.In section 2,we introduce eight conventional non-parametric test statistics and two non-parametric test statistics that apply to the intersection problem..In section 3,we use the numerical simulation to generate different types of right censored survival data.By comparing the test results,we have a comprehensive summary of various non-parametric test statistics.In section 4,three types of real-life data are tested and the results are compared.Finally,section 5 contains a brief summary and discussion.