Analyzing Longitudinal Item Response Data via the Pairwise Fitting Method.

Area of Science:

• Educational Measurement
• Psychometrics
• Statistical Modeling

Background:

• Multidimensional item response theory (MIRT) is applicable to longitudinal educational surveys with common items.
• High-dimensional latent variables in MIRT models present computational challenges, especially for binary data where analytical integration is difficult.

Purpose of the Study:

• To propose a novel pairwise modeling strategy for estimating item and population parameters in longitudinal studies using MIRT.
• To address the computational difficulties associated with high-dimensional latent variables in longitudinal educational data.

Main Methods:

• The study utilizes pseudolikelihood theory to develop a pairwise modeling strategy.
• This approach effectively reduces the dimensionality of the estimation problem.

Main Results:

• The proposed pairwise method is applicable to longitudinal IRT data with high-dimensional latent variables, outperforming classical methods in challenging scenarios.
• Simulations demonstrate comparable performance to classical methods in low-dimensional cases.

Conclusions:

• The pairwise modeling strategy offers a computationally efficient and effective solution for analyzing longitudinal educational data with MIRT.
• The method was successfully illustrated using a study on junior high school mathematics levels in China.