Testing the Nonlinearity Assumption Underlying the Use of Reverse-Keyed Items: A Logical Response Perspective

Area of Science:

• Psychometrics
• Quantitative Psychology
• Survey Methodology

Background:

• Researchers commonly assume a linear relationship between regular- and reverse-keyed survey items.
• This assumption implies responses on one item type perfectly mirror the other.

Purpose of the Study:

• To challenge the assumed linear relationship between regular- and reverse-keyed items.
• To propose and demonstrate a nonlinear relationship model.
• To investigate the impact of item characteristics on this relationship.

Main Methods:

• Analysis of four datasets (N = 50,544) including human height, job satisfaction, affect, and self-esteem.
• Comparison of linear and nonlinear models to explain item variance.
• Examination of moderation effects by item characteristics like extremity and softening.

Main Results:

• A nonlinear model consistently explained additional variance beyond a linear model across all datasets.
• Item characteristics, such as extremity and softening, moderated the nonlinear relationship.
• Mismodeling this relationship can lead to the false conclusion of construct bidimensionality.

Conclusions:

• The relationship between regular- and reverse-keyed items is often nonlinear, not linear.
• Accounting for nonlinearity improves construct measurement accuracy.
• User-friendly syntax is provided to facilitate the analysis of this nonlinear relationship.