There are problems of comparing two or more samples in terms of the risks which arise during the selecting of statistical population in applied statistics and in data analysis. In such problems it is convenient to consider one generalized criterion that makes it possible to estimate the homogeneity of the statistical populations from the point of view of assessing possible risks, when several indicators characterizing the magnitude of the risk are being considered simultaneously. In this paper, we offer a criterion for comparing several samples, when the risk is estimated from sample variances and sample means. Such a comparison belongs to the class of multicriteria problems. The proposed criterion can be applied to samples from statistical populations with different distributions. The criterion includes one configurable parameter, the options for selecting this parameter are proposed in this paper.