Fit analysis in latent trait measurement models.

Area of Science:

• Psychometrics
• Statistical Modeling

Background:

• Latent trait models, including the Rasch model, rely on data-model fit for accurate measurement.
• Desirable model characteristics like interval measures and parameter invariance depend on the data fitting the model.
• Poor data-model fit diminishes the validity of these characteristics.

Purpose of the Study:

• To explore the nature of data-model fit in latent trait models.
• To provide a historical overview of fit indices.
• To demonstrate the necessity of standardized fit indices for comprehensive model evaluation.

Main Methods:

• Review of the concept of fit in latent trait models.
• Historical analysis of various fit indices.
• Focus on Pearsonian chi-square-based fit indices.
• Examination of standardized fit indices.

Main Results:

• Data-model fit is a prerequisite for the validity of Rasch model properties.
• A historical progression of fit indices has been observed.
• A single fit index is insufficient for a complete understanding of model fit.
• Standardized fit indices are essential for a thorough assessment.

Conclusions:

• The interpretation of Rasch model calibration requires rigorous fit analysis.
• A family of standardized fit indices is necessary for a complete understanding of data-model relationships.
• Accurate measurement depends on the appropriate application and interpretation of fit indices.