Joint regression and association modeling of longitudinal ordinal data

Area of Science:

• Statistics
• Biostatistics
• Longitudinal Data Analysis

Background:

• Ordinal data with clustering is common in scientific research.
• Existing methods may not fully capture complex associations in such data.
• Understanding treatment effects in longitudinal studies requires robust statistical approaches.

Purpose of the Study:

• To propose novel statistical models for longitudinal or clustered ordinal data.
• To extend dependence ratios from binary to multicategory cases for characterizing associations.
• To develop regression models for cumulative probabilities and derive dependence ratios from association mechanisms.

Main Methods:

• Utilizing extended dependence ratios for multicategory ordinal data.
• Expressing joint probabilities as functions of marginal means and dependence ratios.
• Implementing likelihood-based inference with computational advantages.
• Applying selection models to assess sensitivity to drop-out.

Main Results:

• The proposed models provide a unified framework for analyzing clustered ordinal data.
• Explicit functions of marginal means and dependence ratios facilitate inference.
• The analysis of Fluvoxamine treatment data demonstrates the models' practical application.
• Sensitivity analysis using selection models addresses potential drop-out biases.

Conclusions:

• The developed models offer a flexible and computationally efficient approach for clustered ordinal data.
• The framework allows for detailed investigation of associations and regression.
• The study provides valuable insights into treatment effects and data attrition in longitudinal research.