Combining evidence from several samples for testing goodness-of-fit to a location-scale family. Technical report No. 15

Consider the problem of testing goodness-of-fit to a specified location-scale family when evidence is to be combined from several independent samples, from populations with possibly different location and scale parameters. The procedure studied is that of computing standardized residuals from each sample and then combining these into one set to be treated essentially as though they came from one sample. It is shown that the limiting distribution of any location-scale invariant goodness-of-fit statistic so applied is precisely the same as for the corresponding one-sample problem. 13 references.