Two-sample sequential test of normal distributions when one of the sample sizes is fixed

Abstract

Sequential two-sample tests when the sample size for one of the two populations is fixed is discussed. Two tests, one for comparing the means and the other for comparing variances of two normal distributions, are presented together with the test boundaries. The numerical approach of determining the risk probabilities by integrating the conditional error probabilities with respect to the distributions of estimators is employed. The proposed tests appear to yield an appreciable advantage over the usual fixed sample size test in terms of the average total number of observations.