Analysis of Incomplete Longitudinal Binary Data-A Combined Markov's Transition and Logistic Model for Non-ignorable...

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Clinical Trial Methodology

Background:

• Incomplete data is a pervasive challenge in longitudinal research.
• Existing methods often struggle with non-ignorable missing data mechanisms.
• Accurate statistical modeling is vital for reliable research outcomes.

Purpose of the Study:

• To develop a likelihood-based statistical approach for analyzing incomplete longitudinal binary data.
• To address scenarios where the missing data mechanism is non-ignorable.
• To provide a robust method applicable to complex clinical trial data.

Main Methods:

• Utilized a disposition model incorporating Markov's transition and logistic regression for dropout processes.
• Employed conditional logistic regression to model the longitudinal binary response.
• Conducted sensitivity analysis by holding weakly identified missingness parameters constant.

Main Results:

• The proposed method offers a viable approach for estimating parameters in the presence of non-ignorable missing data.
• Sensitivity analysis demonstrated the impact of missingness parameters on model estimation.
• The approach was successfully applied to a Schizophrenia clinical trial dataset.

Conclusions:

• The developed likelihood-based method effectively handles non-ignorable incomplete longitudinal binary data.
• This statistical framework enhances the analysis of complex clinical trial data.
• The study provides a valuable tool for researchers dealing with missing data in longitudinal studies.