Asymptotic distributions of some test criteria for the covariance matrix in elliptical distributions under local alternatives (Q1907816)

The objective of this paper is to obtain under local alternatives asymptotic expansions of distributions of test criteria for some special problems, for testing for the covariance matrix of an elliptical distribution. The organization of the paper is as follows. In Section 2, we describe the elliptical model and test statistics for each problem. Section 3 deals with the case of testing that the characteristic roots of the covariance matrix, assumed distinct, are equal to a set of specified numbers. Section 4 deals with the case of testing that the covariance matrix is equal to a specified positive definite matrix, which is assumed to be the identity matrix without loss of generality. We compare the performances of the test criteria for each of the problems in terms of the behaviour of their power functions. Section 5 deals with a modified likelihood ratio criterion obtained under a normal model for testing the covariance structure in familial data, when the observations are indeed from the elliptical distribution described above. In Section 6, we compute the powers of the test statistics studied in Sections 3 and 4, by using the expansions of their distributions obtained in these sections.