Residuals analysis of the generalized linear models for longitudinal data.

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Medical Research Methodology

Background:

• Generalized Estimation Equation (GEE) is a standard method for analyzing longitudinal data in medical research.
• Sensitivity analysis for GEE models, particularly concerning correlated data structures, remains underexplored.
• Conventional residual plots may inaccurately validate models for longitudinal data.

Purpose of the Study:

• To address the gap in sensitivity analysis for GEE models in longitudinal studies.
• To propose a reliable method for diagnosing model fit in the presence of correlated data.
• To enhance the trustworthiness of statistical models used in medical research.

Main Methods:

• Utilized Generalized Estimation Equation (GEE) models for longitudinal data.
• Investigated the limitations of conventional residual plots for model diagnosis.
• Introduced and applied the non-parametric Wald-Wolfowitz run test for quantitative and graphical residual analysis.

Main Results:

• Demonstrated that conventional residual plots can be misleading for longitudinal data.
• The Wald-Wolfowitz run test provides a robust method for evaluating model residuals.
• The proposed method was validated using two real clinical studies.

Conclusions:

• The Wald-Wolfowitz run test is a valuable tool for sensitivity analysis in GEE models.
• Accurate model diagnostics are crucial for reliable medical research findings.
• This approach improves the reliability of statistical modeling for longitudinal health data.