Biostatistics Seminar: Keith Widaman (Dept. Psychology, UC Davis)
DATE: Tuesday, April 2nd, 2013
LOCATION: 1043 Gladys Valley Hall (map)
SPEAKER: Keith Widaman, PhD, Distinguished Professor, Department of Psychology, UC Davis
TITLE: Distinguishing Ordinal and Disordinal Interactions – With Application to Gene X Environment Interaction Studies
ABSTRACT: An ordinal interaction between two predictors, X1 and X2, preserves the rank order of predicted values as a function of X2 across the observed range of X1. In contrast, a disordinal interaction contains a crossover point on X1, such that values of X2 that lead to more positive predicted values above the crossover point are associated with more negative predicted values below the crossover point. The standard approach to specifying linear regression models with a product term (X1 x X2) for the linear-by-linear interaction can be used, but the crossover point is not a parameter in the model and requires additional computation. In recent work on Gene x Environment (GxE) interactions, two theoretical positions have been outlined – Diathesis-Stress and Differential Susceptibility. The key differential prediction for these two opposing positions is the location of the crossover point: Diathesis-Stress predicts an ordinal interaction, whereas Differential Susceptibility predicts a disordinal interaction. We re-parameterized the linear regression model so the crossover point is a parameter to be estimated. This led to a point estimate and an associated SE, allowing an interval estimate of this parameter. Various extensions in the context of GxE interaction research are possible, including (a) more direct tests of other moderators (e.g., participant sex) on key model parameters, and (b) alternate models of gene action (i.e., linear, nonlinear, recessive, dominant). Examples will be provided.