Novel U-empirical moment-generated-function-based tests for exponentiality

Abstract

We present two novel scale-free classes of goodness-of-fit tests for exponentiality based on U-empirical moment-generating functions. The tests are constructed using the Puri–Rubin characterization of the exponential distribution; however, the approach can be applied to any equidistribution characterization. The limiting distributions of the proposed statistics are derived. In an extensive simulation study, the tests are compared with several recent powerful competitors. The good performance of the new tests provides strong motivation for studying broader classes of equidistribution-type characterization-based tests using the considered integral transformation.