New consistent goodness-of-fit tests based on V-empirical Laplace transforms

Abstract

New goodness-of-fit tests are proposed which employ the equidistribution characterizations. Based on V-empirical Laplace transforms of equidistributed statistics, test statistics of L^p-type are constructed. Their asymptotic properties are derived. To assess their quality, the approximate Bahadur efficiency is used. For small sample sizes, a simulated power study is performed. The tests are shown to be very efficient and powerful in comparison to some common competitors.