A random item effects generalized partial credit model with a multiple imputation-based scoring procedure.

Area of Science:

• Psychometrics
• Statistical modeling
• Item response theory

Background:

• Random item effects item response theory (IRT) models are increasingly utilized.
• Further research is needed for polytomous data and scoring procedures.
• Addressing these gaps enhances the utility of advanced IRT models.

Purpose of the Study:

• To propose a novel random item effects generalized partial credit model (GPCM) for polytomous data.
• To introduce a multiple imputation (MI)-based scoring procedure for random item effects IRT models.
• To evaluate the proposed model and scoring method using empirical and simulation data.

Main Methods:

• Developed a new GPCM incorporating random person, item, and category-specific effects.
• Implemented an MI-based scoring procedure applicable to various random item effects IRT models.
• Analyzed Quality of Life (QoL) scale data and conducted a simulation study.

Main Results:

• Patient scores from the proposed model were similar to conventional GPCM scores.
• Standard errors for scores were slightly larger with the proposed approach.
• The simulation study demonstrated adequate recovery of model parameters and patient scores.

Conclusions:

• The proposed GPCM and MI scoring procedure advance IRT methodology.
• The GPCM reduces free parameters, beneficial for small sample sizes.
• The MI scoring procedure effectively addresses scoring issues and is extensible.