Finite Mixtures for Simultaneously Modelling Differential Effects and Non-Normal Distributions

Area of Science:

• Social and behavioral sciences
• Statistical modeling
• Quantitative psychology

Background:

• Regression mixture models identify differential predictor effects.
• Standard models are sensitive to non-normal error distributions.
• Existing methods for capturing differential effects need improvement.

Purpose of the Study:

• Introduce and test a novel approach using differential effects sets.
• Simultaneously model differential effects and account for non-normal error distributions.
• Improve the description of differential effects in regression mixture models.

Main Methods:

• Utilized Monte Carlo simulations to evaluate the new approach.
• Employed differential effects sets to model non-normal error distributions.
• Applied the method to analyze parental health problems and adolescent BMI.

Main Results:

• The number of classes required depends on the degree of skewness.
• Differential effects sets reduced bias in parameter estimates.
• Demonstrated practical implementation in an applied analysis.

Conclusions:

• The differential effects sets approach effectively handles non-normal errors.
• This method overcomes limitations of previous regression mixture models.
• Supports the utility of the approach for describing complex relationships.