Bayesian tests of measurement invariance.

Area of Science:

• Psychometrics
• Statistical Modeling

Background:

• Measurement invariance is crucial for cross-national comparisons.
• Traditional methods often require anchor items, limiting their application.
• Random item effects models offer an alternative framework.

Purpose of the Study:

• To propose and evaluate Bayesian tests for measurement invariance within a random item effects framework.
• To enable simultaneous testing of multiple invariance hypotheses.
• To explore cross-national variation in item functioning.

Main Methods:

• Utilizing random item effects models.
• Implementing Bayesian statistical tests, including Bayes factor and deviance information criterion.
• Conducting a simulation study to assess test performance.
• Analyzing data from the European Social Survey.

Main Results:

• The proposed Bayesian tests demonstrate high statistical power and low Type I error rates.
• Multiple marginal invariance hypotheses can be tested concurrently.
• Background information can explain variations in item functioning across countries.

Conclusions:

• Random item effects models provide a robust framework for assessing measurement invariance.
• Bayesian tests within this framework are effective for identifying cross-national differences in item bias.
• This approach enhances the validity of cross-national survey research.