Two item response theory models for analysing normative forced-choice personality items.

Area of Science:

• Psychometrics
• Personality Psychology
• Statistical Modeling

Background:

• Item response theory (IRT) is a key statistical framework for analyzing test and questionnaire data.
• Forced-choice personality items, common in assessments, present unique analytical challenges.
• Existing IRT models may not fully capture the nuances of forced-choice response mechanisms.

Purpose of the Study:

• To propose and evaluate two novel unidimensional item response theory (IRT) models specifically designed for normative forced-choice personality items.
• To explore different assumptions about choice mechanisms to derive distinct IRT models.
• To compare the theoretical and empirical properties of the proposed models.

Main Methods:

• Development of two unidimensional IRT models from a common theoretical framework based on differing choice mechanism assumptions.
• Theoretical comparison of the proposed models, including the normal-ogive and a new model related to the Coombs-Zinnes probabilistic unfolding model, focusing on item characteristic curves and item information.
• Empirical comparison of the two models using data from established personality measures.
• Description of parameter estimation procedures for the novel IRT model.

Main Results:

• The simplest choice mechanism leads to the one-parameter normal-ogive model.
• A more complex mechanism results in a new IRT model, closely aligned with the Coombs-Zinnes probabilistic unfolding model.
• Theoretical analysis highlights differences in item characteristic curves and item information between the models.
• Empirical evaluation using real personality data provides comparative insights into model performance.

Conclusions:

• The proposed IRT models offer viable alternatives for analyzing forced-choice personality data.
• The choice mechanism assumptions significantly influence the resulting IRT model properties.
• These models enhance the psychometric analysis of personality questionnaires, potentially leading to more accurate trait measurement.