New characterization based goodness-of-fit tests for randomly censored data

Abstract

Recently, the characterization based approach for the construction of goodness of fit tests has become popular. Most of the proposed tests have been designed for complete i.i.d. samples. We present the adaptation of the recently proposed exponentiality tests based on equidistribution-type characterizations for the case of randomly censored data. Their asymptotic properties are provided. We also present the results of a wide empirical power study, including the powers of several recent competitors. The results can be used as a benchmark for future tests proposed for this kind of data.