Lee Van Horn
Department of Psychology
University of South Carolina
Moderation Without a Moderator:
Assessing Regression Mixture Models
This talk with present preliminary findings from a study evaluating the use
of finite mixture models, where the mixture is defined by differential
regression weights, to assess moderation. Regression mixture models are not
themselves new, but they have received little attention in the
methodological or applied literatures, perhaps because they require large
sample sizes and are highly reliant on distributional assumptions. This
research aims to better understand their potential use and the implications
of violating the assumptions. Results using both real and simulated data
will be presented.
Back to Colloquium Series