Generalized estimating equations to estimate the ordered stereotype logit model for panel data

Area of Science:

• Statistics
• Econometrics
• Biostatistics

Background:

• Ordered stereotype logit models offer flexibility for ordinal response variables.
• Panel data analysis requires robust methods to handle correlated observations over time.
• Generalized estimating equations (GEE) provide a powerful framework for analyzing correlated data.

Purpose of the Study:

• To propose and evaluate a GEE approach for estimating ordered stereotype logit models with panel data.
• To compare the efficiency of different working covariance structures within the GEE framework.
• To assess the feasibility and performance of GEE estimation in various correlation scenarios.

Main Methods:

• Development of a GEE approach for ordered stereotype logit models.
• Simulation studies comparing GEE estimators with different working matrices (correlation and covariance).
• Analysis of a real-world panel dataset to illustrate model estimation.

Main Results:

• GEE estimation is feasible and reliable for medium to large sample sizes.
• Working correlation matrices generally yield more efficient GEE estimators than local odds ratios.
• High true correlations benefit significantly from working covariance structures that closely match the true correlation.

Conclusions:

• The proposed GEE approach is a viable method for analyzing ordered stereotype logit models with panel data.
• Careful selection of the working covariance structure is crucial for maximizing efficiency, especially with high correlations.
• GEE offers a flexible and robust alternative to traditional methods for longitudinal ordinal data analysis.