Article,

Robust Procedures for Testing Equality of Covariance Matrices

.
Biometrics, 48 (3): pp. 819-827 (1992)

Abstract

Normal-theory tests for common variance are very sensitive to departures from normality in both univariate and multivariate applications. For mild departures from normality and moderate sample size, nominal α = .01 produced Type I error rates in excess of .20 in simulation studies. Since a modification of Levene's test has been shown to perform well for the univariate problem, two multivariate generalizations of this test are proposed. Both procedures are robust to departures from normality, and are simple computationally. Comparisons of the power of these two procedures and a procedure proposed previously by Tiku and Balakrishnan (1985, Communications in Statistics-Theory and Methods 14, 3033-3051) indicate that the three methods should be viewed as complementary.

Tags

Users

  • @vivion

Comments and Reviews