O porównywaniu dwóch populacji wielowymiarowych z wykorzystaniem objętości elipsoid ufności

A comparison of two populations seems to be interesting and very common statistical problem. The most often way is to verify the hypothesis concerned the equality of certain, characteristic parameter i.e. mean, standard deviation or fraction with parametric or non-parametric tests. The authors propose to compare the distribution of two populations - comparing the confidence ellipsoid volumes. Since their distribution is unknown - permutation tests were applied. A Monte-Carlo simulation let to compare power of these tests with T2 Hotelling tests. Proposed methods can be used, when the assumptions for parametric tests couldn't be verified.