Comparison of Tests for Heteroscedasticity in Linear Models

• In Event: Linear Modeling: Methodology and Applications

Abstract

In some research applications, it is of scientific interest to investigate whether the variability of an outcome is affected by an experimental treatment or changes over time. Several tests for differences in variances in basic between-subjects designs have been developed. These procedures transform the original response, y, then perform one-way ANOVA on the transformed data. Breusch and Pagan (1979) generalized the Levene (1960) approach of using squared residuals to linear regression models. Variance Function Regression (VFR) uses a generalized linear gamma model with the squared residuals as the response. A simulation study will investigate how these methods for testing variances perform in terms of test size and power for three models: between-subjects ANOVA, repeated measures, and general regression.

Author

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