A comprehensive empirical power comparison of univariate goodness-of-fit tests for the Laplace distribution
share
In this paper, we do a comprehensive survey of all univariate goodness-of-fit tests that we could find in the literature for the Laplace distribution, which amounts to a total of 45 different test statistics. After eliminating duplicates and considering parameters that yield the best power for each test, we obtain a total of 38 different test statistics. An empirical power comparison study of unmatched size is then conducted using Monte Carlo simulations, with 400 alternatives spanning over 20 families of distributions, for various sample sizes and confidence levels. A discussion of the results follows, where the best tests are selected for different classes of alternatives. A similar study was conducted for the normal distribution in Romão et al. (2010), although on a smaller scale. Our work improves significantly on Puig Stephens (2000), which was previously the best-known reference of this kind for the Laplace distribution. All test statistics and alternatives considered here are integrated within the PoweR package for the R software.
READ FULL TEXT
