A Multidimensional Continuous Response Model for Measuring Unipolar Traits.

Area of Science:

• Psychometrics
• Item Response Theory (IRT)

Background:

• Unipolar constructs are prevalent in various non-cognitive assessments, including clinical, forensic, and symptom checklists.
• Existing Item Response Theory (IRT) models for unipolar constructs, like the Log-Logistic model, are limited to unidimensional structures.

Purpose of the Study:

• To propose a novel multidimensional log-logistic Item Response Theory (IRT) model for double-bounded continuous response items measuring unipolar constructs.
• To address the limitations of current unidimensional IRT models in analyzing complex, multidimensional unipolar measures.

Main Methods:

• Development of a multidimensional log-logistic IRT model suitable for continuous response data.
• Introduction of multidimensional item location and discrimination indices.
• Implementation of procedures for model fitting, respondent scoring, and accuracy assessment using R software.

Main Results:

• The proposed model demonstrates simplicity and useful properties, adaptable through linearizing transformations.
• Empirical validation using data from 371 students on the Brief Symptom Inventory and Rosenberg Self-Esteem Scale.
• The model effectively interprets unipolar variables, improving conditional reliability of trait estimates and external validity.

Conclusions:

• The novel multidimensional log-logistic IRT model offers a valuable tool for analyzing unipolar constructs in multidimensional contexts.
• The model enhances the accuracy and interpretability of trait estimates in psychological and behavioral assessments.
• The R implementation ensures practical application and accessibility for researchers in psychometrics and related fields.