A Gibbs sampler for the multidimensional four-parameter logistic item response model via a data augmentation scheme.

Area of Science:

• Psychometrics
• Statistical Modeling

Background:

• The four-parameter logistic (4PL) model is increasingly used for its ability to model the upper asymptote of correct response probability.
• Existing estimation methods for multidimensional 4PL models can be computationally intensive or less accurate.

Purpose of the Study:

• To propose a novel and efficient Gibbs sampling algorithm for estimating the multidimensional four-parameter logistic (4PL) item response model.
• To demonstrate the improved performance of the proposed algorithm compared to existing methods.

Main Methods:

• Development of a Gibbs sampling algorithm utilizing a data augmentation scheme (DAGS).
• Introduction of three continuous latent variables to ensure tractable full conditional distributions.
• Evaluation through simulation studies comparing the proposed method with alternatives.
• Application to an empirical dataset.

Main Results:

• The proposed Gibbs sampling algorithm provides accurate and efficient estimation for the multidimensional 4PL model.
• Simulation studies indicate superior performance of the proposed method over popular alternatives.
• The 4PL model demonstrated improved performance over 3PL and 2PL models on an empirical dataset.

Conclusions:

• The developed Gibbs sampling algorithm offers a practical and effective solution for estimating multidimensional 4PL models.
• The proposed method is accessible to practitioners via the open-source IRTlogit package.
• This advancement facilitates more accurate psychometric analysis in various practical applications.