B0351
Title: A likelihood ratio test for independence and a test of fit for the multivariate linear model for high-dimensional data
Authors: Carlos Coelho - NOVA University of Lisbon, NOVA-Math and NOVA.id.FCT (Portugal) [presenting]
Abstract: A likelihood ratio test (LRT) that may be used for the test of independence of two sets of variables in the high-dimensional case is developed. The main aim is that this LRT may be used as a test of fit for the Multivariate Linear Model and the multi-way MANOVA model, in cases where the number of response variables in the analysis is larger than the number of observations, even allowing for the cells to have just one observation, while enabling the user to test for the significance of individual factors and interactions. The aim is also that the test might be used with non-normal and even heavy-tailed and skewed distributions. A Normal asymptotic distribution is obtained for the test statistic and a test for nested models is also obtained.
