Power Maps in Goodness-of-fit Testing Based on Censored Samples

Abstract

In goodness-of-fit testing based on Type-II right censored samples, mainly either classical tests are applied to transformed samples, which can be considered as full samples under the null hypothesis, or modifications of the tests are applied to the original samples. Results of an extensive simulation study are presented and discussed in order to compare the different approaches and tests by means of powers against beta alternatives. Depending on the beta parameters, illustrative plots lead to recommended tests in given situations, where there is some prior belief about the size of these parameters in the alternative. Moreover, the effects of adding artificial realizations via random dilation are examined.

Keywords:

• Goodness-of-fit tests
• Monte Carlo simulation
• Power study
• Transformation of order statistics
• Type-II censoring.

Mathematics Subject Classification:

• 62G10
• 62G30
• 62N01