Generalized estimating equations for ordinal categorical data: arbitrary patterns of missing responses and...

Area of Science:

• Statistics
• Biostatistics
• Medical Informatics

Background:

• Repeatedly measured ordinal categorical data present analytical challenges, especially with nonmonotone missingness.
• Missing covariates that depend on observable factors introduce potential bias in regression analyses.
• Accurate analysis is crucial in fields like diagnostic oncology for reliable disease status assessment.

Purpose of the Study:

• To develop robust regression methods for ordinal data with nonmonotone missing responses.
• To extend generalized estimating equations (GEEs) to handle missing covariates dependent on observables.
• To correct for potential bias in diagnostic oncology studies with missing true disease status.

Main Methods:

• Utilized ordinal regression models combined with generalized estimating equations (GEEs).
• Extended GEE methodology to accommodate arbitrary missingness patterns independent of unobserved responses.
• Developed bias correction for missing covariates dependent on observable factors.

Main Results:

• The proposed methods effectively handle nonmonotone missingness in ordinal responses.
• The extended GEE framework accommodates complex missing data scenarios.
• Bias correction successfully addresses missing covariate issues, improving analysis accuracy.

Conclusions:

• The developed methods provide a robust framework for analyzing complex ordinal categorical data.
• These techniques are particularly valuable for improving the reliability of diagnostic oncology studies.
• The approach offers a significant advancement in handling missing data in longitudinal studies.