This article presents an asymmetric version of the two-sample t test which
is adjusted for distribution skewness and that is sensitive to alternatives
where one of the population variances may have increased. The need for suc
h a statistic has arisen in testing for parity of service.