Abstract:
|
Recently, there has been a focus on the performance of Heteroskedastic Consistent Covariance Matrix (HCCM) estimators in complex regression models. To date, the performance of HCCM estimators in traditional Analysis of Covariance (ANCOVA) designs has not been directly addressed. We will simulate several heteroskedastic scenarios under the following model: Y = ?0 + ?GG + ?XX + ?, where Y represents the outcome of interests, G is a coding variable for two-group membership and X is an orthogonal fixed effect covariate. Based on simulation conditions from previous research, we will examine the Type 1 and 2 error rates of OLS and HCCM estimators for detecting adjusted mean differences and covariate effects when heteroskedasticity is function of differences in group variances, the covariate (X), and both processes in unbalanced ANCOVA models. Preliminary simulations indicate that under complete and partial null models, heteroskedasticity due to the orthogonal X alone will not affect the Type 1 error rate for the OLS test of H0: ?G = 0; however, heteroscedasticity due to both X and G, may attenuate, exacerbate, or reverse the known effects of heteroskedasticity in unbalanced models.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.