suppose that a treatment effect increases both the mean and the standard deviation of a measurement. can a hypothesis test with z be conducted?