Simulation study of two-sample Kolmogorov-Smirnov test in randomly censored data

Abstract

W artykule przedstawione są trzy wersje testu zgodności Kołmogorowa-Smimowa dla danych prawostronnie cenzurowanych. Poszczególne testy różnią się sposobem podejścia do obserwacji cenzurowanych. Moc testów została zbadana i porównana za pomocą symulacji Monte Carlo.

The paper deals with a problem of testing the non-parametric hypothesis that two populations are equally distributed in the situation when the observations are subject to random censoring. A general metric for measuring the distance between two distributions is the Kolmogorov metric and the corresponding test is the Two-Sample Kolmogorov-Smirnov test. In the report below we present results of a simulation study performed for three versions of the Two-Sample Kolmogorov-Smirnov test for censored data. These three versions are generated by three methods of treating censored observations. Basic statistical properties of these tests are inspected by means of Monte Carlo simulations.