Quantile regression in the presence of monotone missingness with sensitivity analysis

Area of Science:

• Biostatistics
• Statistical Modeling
• Longitudinal Data Analysis

Background:

• Missing data is a common challenge in longitudinal studies.
• Standard statistical methods may produce biased results with incomplete datasets.
• Longitudinal quantile regression is crucial for understanding data distribution over time.

Purpose of the Study:

• To develop robust methods for longitudinal quantile regression with monotone missingness.
• To propose pattern mixture models offering interpretable marginal quantile regression parameters.
• To enable sensitivity analysis for inference with incomplete longitudinal data.

Main Methods:

• Development of pattern mixture models with a specific constraint for interpretable parameters.
• Novel analytic integration techniques to simplify likelihood computation.
• Sensitivity analysis integrated into the inference framework for missing data.

Main Results:

• The proposed methods provide a straightforward interpretation of marginal quantile regression parameters.
• Novel analytic forms for integrals facilitate computational efficiency.
• Simulations demonstrate robustness to modeling assumptions and favorable performance compared to existing methods.

Conclusions:

• The developed methods offer a reliable approach for longitudinal quantile regression with monotone missing data.
• The technique is applicable to real-world data, such as clinical trial results for weight management.
• The study enhances statistical inference capabilities for incomplete longitudinal datasets.