Distribution of the largest root of a matrix for Roy's test in
multivariate analysis of variance

Abstract
Let denote two independent real Gaussian and matrices with , each constituted by zero mean i.i.d. columns with common covariance. The Roy's largest root criterion, used in multivariate analysis of variance (MANOVA), is based on the statistic of the largest eigenvalue, , of , where and are independent central Wishart matrices. We derive a new expression and efficient recursive formulas for the exact distribution of . The expression can be easily calculated even for large parameters, eliminating the need of pre-calculated tables for the application of the Roy's test.
View on arXivComments on this paper