Estimation of a four-parameter item response theory model.

Area of Science:

• Psychometrics
• Educational Measurement
• Psychological Assessment

Background:

• Item response theory (IRT) models are crucial for understanding measurement properties.
• Traditional IRT models (2PM, 3PM) may not fully capture response processes.
• Accurate parameter estimation is vital for valid test interpretations.

Purpose of the Study:

• To explore the justification and formulation of the four-parameter item response theory model (4PM).
• To investigate the utility of the 4PM for improving model fit and parameter estimation.
• To examine the impact of using the 4PM versus simpler models on trait score inferences.

Main Methods:

• Bayesian estimation approach for recovering item and respondent parameters.
• Simulation studies using data generated from a 4PM.
• Analysis of an empirical dataset using a widely used delinquency scale.

Main Results:

• The 4PM demonstrated improved overall model fit compared to the 2-parameter and 3-parameter models.
• While trait scores correlated highly across models, the 4PM provided better confidence interval coverage at trait extremes.
• An empirical example showed the 4PM yielding novel insights into scale properties.

Conclusions:

• The 4PM provides a more accurate representation of response processes in certain contexts.
• Using the appropriate IRT model is critical for valid inferences, particularly at the tails of the trait distribution.
• Findings have implications for developing robust measurement models in education and psychology.