Online Program

Visualization Techniques to Help Determine Whether the Continuous Interaction Effect Assumption Is Met Within the Standard Interaction Model (304207)

Keywords: visualization, interaction, non-linearity, continuousness, power

Within the social sciences, it is common to assume the standard interaction model when examining interactions between two continuous predictors. However, methods to assess the assumption of continuous interaction are sparse. Without assessment, discrete changes in the predictor-outcome relationship across regions of the second predictor may be mistaken for continuous interaction. If the standard interaction model is incorrectly specified, estimates may be inaccurate. The present study explored visualizations that may help assess the continuous interaction assumption before the standard interaction model is specified. Visualizations to identify models (continuous interaction, discrete change) were developed by theory. Monte Carlo simulation was conducted to compare models. Power and interaction magnitude were measured overall, then within increasing numbers of regions, as defined by stratifying the second predictor. Simulation results were entered into visualization plots and assessed. Between models, visualizations of type I error displayed similarities (smooth change), while visualizations of power and magnitudes revealed differences (smooth versus oscillating change).