A frailty model for interaction between multiple events.

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Traditional categorical data models often limit interactions to second order.
• High-order interactions can arise unexpectedly from second-order terms, complicating analysis.
• Existing models like the Rasch model struggle with parameter interpretability, especially concerning population heterogeneity ('frailty').

Purpose of the Study:

• To develop a more interpretable model for categorical data exhibiting unavoidable high-order interactions.
• To address limitations of existing models when dealing with population heterogeneity and low event probabilities.
• To propose a multiplicative model with a one-parameter gamma frailty component.

Main Methods:

• Development of a novel multiplicative model for categorical data.
• Implementation of a one-parameter gamma frailty model to account for population heterogeneity.
• Provision of moment estimators for parameter estimation in general cases.
• Inclusion of goodness-of-fit statistics for model validation.

Main Results:

• The proposed gamma frailty model effectively accommodates data where second-order interactions lead to high-order effects.
• The multiplicative structure offers a simpler interpretation compared to logistic models like the Rasch model, especially for low event probabilities.
• Moment estimators and goodness-of-fit statistics are provided for practical application.

Conclusions:

• The developed multiplicative gamma frailty model provides a valuable alternative for analyzing categorical data with complex interaction structures.
• This approach enhances the interpretability of population heterogeneity in statistical modeling.
• The model is demonstrated to be applicable to real-world epidemiological data, such as human papillomavirus prevalence.