The Impact of Different Response Styles on Statistical Analyses for Latent Variable Models

• In Event: Marriott Roundtable Session 9

Abstract

This research quantifies the contaminating effects of various response styles (RSs) on measures of model fit and parameter estimates for latent variable models of increasing levels of complexity. In total, 8 distinct RSs (extreme-avoiding, exclusively-extreme, point-inflation extreme, single-value, non-attending, random, acquiescing, and dis-acquiescing) and 5 select pairs of RSs were examined. A simulation study examines the effects resulting from percent contamination, degree of model complexity, and type of RS contamination. Initial results indicate that effects from extreme RSs may be negligible, effects from acquiescing RSs may be small to moderate, and effects from random and non-attending RSs are large. Impact on different statistical analyses was observed, and impact does appear to be moderated by degree of model complexity.

Authors

• Allen G. Harbaugh-Schattenkirk, Longwood University

• Min Liu, University of Hawaii - Manoa