Performance of Degrees of Freedom Adjustment and Heteroscedastic Consistent Covariance Matrix Methods in Heteroscedastic Factorial ANOVA Models

• In Event: 67.048 - Multiple Linear Regression SIG: The General Linear Model

Abstract

Heterogeneity of variance in one-way ANOVA design has been frequently studied. Zimmerman (2004) concluded a separate-variances test should be used “unconditionally whenever sample sizes are unequal.” However, has been little attention to heteroscedasticity in Factorial ANOVA (FANOVA). Statisticians at SAS (Littell et al., 2006) have suggested the use of approximate degrees-of-freedom (ADF) tests based on Kenward & Roger (1997). Recently, there has also been a focus on the performance of Heteroscedastic Consistent Covariance Matrix (HCCM) estimators in complex regression models. To date, the performance of these approaches in FANOVA models has not been directly addressed. The Type 1 and 2 error rates of the methods for detecting main effects and interactions under heteroscedasticity were compared.

Author

• T. Mark Beasley, The University of Alabama - Birmingham