Multivariate symmetry via projection pursuit (Q753351)

A measure of symmetry of multivariate distributions is defined and then used as an index for projection pursuit to find ``directions of symmetry''. For finding such directions one may use on (one-dimensional) projected data a test of symmetry. In the case when the hypothesis of symmetry fails to be rejected one may restrict himself/herself on the orthogonal subspace (orthogonal to the found ``direction of symmetry'') and the whole procedure may be repeated. Exhausting all dimensions of data one finds centrosymmetry. In the opposite case one finds the degree of symmetry. Illustration by real data (concerning an entomologist experiment on cotton fields of Arizona) is included.