A Rating Scale Mixture Model to Account for the Tendency to Middle and Extreme Categories

• Published in: Journal of educational and behavioral statistics Vol. 46; no. 6; pp. 682 - 716
• Main Authors: Colombi, Roberto, Giordano, Sabrina, Tutz, Gerhard
• Format: Journal Article
• Language: English
• Published: Los Angeles, CA SAGE Publications 01-12-2021; American Educational Research Association
• Subjects: Attitude Measures; Questionnaires; Rating Scales; Regression (Statistics); Response Style (Tests); Test Items; marginal models; latent variables; response styles; logit models
• ISSN: 1076-9986; 1935-1054
• DOI: 10.3102/1076998621992554

A mixture of logit models is proposed that discriminates between responses to rating questions that are affected by a tendency to prefer middle or extremes of the scale regardless of the content of the item (response styles) and purely content-driven preferences. Explanatory variables are used to characterize the content-driven way of answering as well as the tendency to middle or extreme categories. The proposed model is extended to account for the presence of response styles in the case of several items, and the association among responses is described, both when they are content driven or dictated by response styles. In addition, stochastic orderings, related to the tendency to select middle or extreme categories, are introduced and investigated. A simulation study describes the effectiveness of the proposed model, and an application to a questionnaire on attitudes toward ethnic minorities illustrates the applicability of the modeling approach.