Using Factor Mixture Models to Identify Response Styles Associated With Negatively Keyed Items

• In Event: Response Styles and Measurement Invariance in Factor Modeling

Abstract

The purpose of this study is to demonstrate the utility of factor mixture models for detecting latent classes based on differential response patterns to positively- and negatively-keyed items using three attitude/affect measures. Numerous studies have found that balanced/mixed scales yield “artifactors” or method effects that are defined by item keying/ wording direction. Posited explanations for this phenomenon include the presence of subgroups of individuals who respond in a careless, acquiescent, or otherwise aberrant manner, leading to differential response distributions for negative and positive items. Factor mixture models provide the flexibility to detect method effects that may affect some respondents but not others.

Authors

• Jerusha J. Gerstner, James Madison University

• Chris M. Coleman, Babson College

• Deborah L. Bandalos, James Madison University