New characterization based exponentiality tests for randomly censored data

TL;DR

This paper adapts exponentiality goodness-of-fit tests based on characterizations to handle randomly censored data, providing asymptotic properties and empirical power comparisons.

Contribution

It introduces new exponentiality tests for censored data using characterization methods, with proven asymptotic properties and extensive power analysis.

Findings

Tests have good asymptotic properties.

Empirical power studies show competitive performance.

Provides a benchmark for future censored data tests.

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. Here 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. Besides, we present the results of wide empirical power study including the powers of several recent competitors. This study can be used as a benchmark for future tests proposed for this kind of data.