Modeling Qualitative Variation Within Latent Trait Dimensions: Application of Mixed-Measurement to Personality...

Area of Science:

• Psychometrics
• Statistical Modeling

Background:

• Item response theory (IRT) and factor analytic models offer distinct frameworks for analyzing item response data.
• Understanding the underlying structure of item response matrices is crucial for accurate data interpretation.

Purpose of the Study:

• To contrast mixed-measurement IRT models and traditional factor analytic models.
• To determine the most appropriate model for representing data from a measure of Positive Interpersonal Engagement.
• To explore statistical and conceptual criteria for model selection.

Main Methods:

• Estimated parameters for full-information item factor models with varying dimensionality.
• Estimated parameters for mixed-measurement models with varying numbers of latent classes.
• Utilized a sample of 1,000 responses to a 14-item Positive Interpersonal Engagement measure.

Main Results:

• Both two latent factor and two latent class mixed-models provided appropriate data representations.
• Factor models generally demonstrated greater parsimony compared to mixed-measurement models.
• Statistical fit alone is insufficient for choosing between dimensional and mixed-measurement models.

Conclusions:

• The choice between dimensional (factor) and mixed-measurement (latent class) models should consider conceptual appropriateness beyond statistical criteria.
• Mixed-measurement models may be conceptually superior in specific research contexts.
• Both approaches offer valuable insights into item response heterogeneity.