A local sensitivity analysis approach to longitudinal non-Gaussian data with non-ignorable dropout

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Missing Data Methods

Background:

• Analyzing longitudinal non-Gaussian data with non-ignorable dropout presents significant statistical challenges.
• Standard methods often rely on unverifiable ignorability assumptions for missing data.
• Sensitivity analysis is crucial for assessing the robustness of results to alternative dropout assumptions.

Purpose of the Study:

• To extend the index of local sensitivity to non-ignorability (ISNI) methodology.
• To quantify the sensitivity of inferences for longitudinal non-Gaussian data with non-ignorable dropout.
• To evaluate the proposed methodology using simulations and a real-world example.

Main Methods:

• Extension of the ISNI methodology to generalized linear mixed models for longitudinal data.
• Quantification of sensitivity in the neighborhood of a missing at random (MAR) model.
• Simulation studies to assess performance.
• Application to smoking-cessation data.

Main Results:

• The proposed ISNI methodology effectively quantifies sensitivity to non-ignorable dropout in longitudinal non-Gaussian settings.
• Simulation results demonstrate the practical utility and performance of the extended method.
• The real-world example illustrates the application and interpretation of the sensitivity analysis.

Conclusions:

• The extended ISNI provides a valuable tool for assessing the impact of non-ignorable dropout on longitudinal non-Gaussian data.
• Researchers can use this method to evaluate the robustness of their findings under various dropout scenarios.
• This approach enhances the reliability of statistical inferences in the presence of complex missing data patterns.