Estimation of Latent Variable Scores with Multiple Group Item Response Models: Implications for Integrative Data...

Area of Science:

• Psychometrics
• Data Science
• Statistical Modeling

Background:

• Integrative data analysis (IDA) combines multiple datasets for joint analysis.
• Scaling multisample item-level data to a common metric is crucial for IDA.
• Multiple group item response models (MGM) are commonly used for this scaling step.

Purpose of the Study:

• To evaluate the accuracy of latent variable estimates from MGM compared to alternative methods.
• To investigate the impact of different approaches on latent variable estimation in IDA.
• To provide guidance on appropriate methods for data scaling in IDA.

Main Methods:

• Utilized multiple group item response models (MGM) with invariance constraints.
• Conducted a Monte Carlo simulation study to compare estimation methods.
• Compared MGM, single-group item response models, and MGM ignoring group differences.

Main Results:

• Alternative approaches, including ignoring group differences, yielded consistent latent variable estimates.
• The accuracy of latent variable estimates was comparable across tested methods.
• MGM with invariance constraints did not consistently outperform simpler alternatives.

Conclusions:

• Simpler methods may be sufficient for latent variable estimation in IDA, challenging the necessity of complex MGM.
• The choice of scaling method may have less impact on latent variable accuracy than previously assumed.
• Findings suggest potential simplification of data preparation steps in integrative data analysis.