Item response mixture modeling: application to tobacco dependence criteria

Area of Science:

• Psychometrics
• Statistical Modeling
• Behavioral Science

Background:

• Conventional methods like factor analysis and latent class analysis have limitations in phenotypic analysis.
• Analyzing categorical data within hybrid models presents unique statistical challenges.
• Recent advancements have made hybrid model analysis for categorical data more feasible.

Purpose of the Study:

• To introduce and illustrate novel hybrid latent variable models for phenotypic analyses.
• To demonstrate the application of these models to categorical item analysis.
• To compare the performance of hybrid models against traditional factor analysis and latent class analysis.

Main Methods:

• Development of hybrid latent variable models combining dimensional and categorical features.
• Application of Mplus software for practical implementation of hybrid models with categorical data.
• Comparative analysis of model fit between hybrid and conventional statistical techniques.

Main Results:

• Hybrid latent variable models demonstrate a superior fit to data compared to traditional factor analysis and latent class analysis.
• The models are shown to be practical for analyzing complex categorical item data.
• Successful illustration of hybrid model application in a tobacco dependence survey analysis.

Conclusions:

• Hybrid latent variable models represent a promising advancement for phenotypic research.
• These models offer enhanced analytical capabilities, particularly for categorical data.
• The findings support the utility of hybrid models for better understanding complex behavioral phenomena.