Title: Behaviour of tests in the Cox proportional hazards model
Authors: Aneta Andrasikova - Palacky University Olomouc (Czech Republic) [presenting]
Eva Fiserova - Palacky University (Czech Republic)
Abstract: Survival analysis, or so called time-to-event analysis, is applied in a wide range of research fields, especially in medicine, where it allows evaluation of certain medical procedures. It is based on evaluating the time until the occurrence of the event of interest, e.g. relapse of certain disease or death. Typical features of survival data is censoring, i.e. incomplete information about the time of occurrence of the event. This incompleteness may be caused by the end of the study before occurrence of the event or loss of contact with subject of observation. These observations are called as right-censored. The effect of some specific covariates on survival time is usually analysed by the Cox proportional hazards model and the statistical significance of the effect of covariates is then verified by the likelihood ratio test, the Wald test, or the score test. These tests are asymptotically equivalent; however, they give numerically different results in applications in dependence on available data. The author deals with the behaviour of the power of these tests for small sample sizes under various proportion of right censored data and distributions of baseline hazard functions. The aim is to familiarize the audience with the results of performed simulations.