what would unambiguously increase the value of the test statistic?
is it an increase in the sample variance for one treatment groups whose variances were previously below the combined sample variance?