Testing for scale families using total variation distance
Sneh Gulati, Jie Mi · Journal of Statistical Computation and Simulation · 2006
Total variation distance is used to make inferences about scale families. We examine the distribution of the test statistic for both large and small samples and show that, for large samples, the test statistic converges to a half normal distribution. The procedure can also be used to test whether two independent samples come from identical distributions vs. the hypothesis that they come from different distributions. The one-sample test is shown to be applicable to a number of families. In particular, tests for the Weibull distribution and the gamma distribution are developed as a special application of the single-sample case.