When two samples have been taken from normal distributions that have the same (but unknown) variance and possibly different means, then a t-test (see hypothesis test) is appropriate. In other cases a non-parametric test may be used (see test for equality of location). See also Behrens–Fisher problem; Welch statistic.