An AIC-type information criterion evaluating theory-based hypotheses for contingency tables

Area of Science:

• Statistics
• Data Analysis
• Statistical Modeling

Background:

• Evaluating theory-based hypotheses in contingency tables presents challenges.
• Log-linear models often lack the capacity to handle complex relationships and inequality restrictions.
• High-dimensional tables with sparse or empty cells complicate parameter estimation and interpretation.

Purpose of the Study:

• To propose a novel method simplifying the evaluation of theory-based hypotheses in high-dimensional contingency tables.
• To address limitations of traditional log-linear models, including handling inequality constraints and sparse data.
• To introduce an AIC-type information criterion, GORICA, for evaluating hypotheses with mixed constraints.

Main Methods:

• Development of the GORICA (Generalized Order Restricted Information Criterion for Analysis) method.
• Specification of theory-based hypotheses using equality and/or inequality constraints on cell probabilities.
• Evaluation of GORICA's performance through a simulation study on contingency tables.
• Application of the method to two empirical examples.

Main Results:

• The proposed method effectively simplifies the evaluation of complex hypotheses in high-dimensional contingency tables.
• GORICA demonstrates robust performance in simulation studies, handling sparse data and mixed constraints.
• The method enhances interpretability compared to traditional log-linear approaches.

Conclusions:

• GORICA offers a powerful and flexible alternative for analyzing theory-based hypotheses in challenging contingency table data.
• The method successfully integrates equality and inequality constraints, improving upon existing statistical approaches.
• Empirical examples showcase the practical utility and effectiveness of GORICA in real-world research.