Likelihood-based analysis of longitudinal data from outcome-related sampling designs

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Statistical Modeling

Background:

• Longitudinal data analysis is crucial for understanding changes over time.
• Rare or expensive outcomes necessitate efficient sampling strategies.
• Outcome-related sampling can enhance data collection for specific research questions.

Purpose of the Study:

• To develop novel likelihood-based methods for generalized linear mixed models (GLMMs).
• To accommodate diverse outcome-related sampling designs in longitudinal studies.
• To improve the efficiency and reduce the cost of analyzing longitudinal data with challenging outcomes.

Main Methods:

• Developed two likelihood-based approaches for fitting GLMMs.
• Extended a semi-parametric maximum likelihood approach for general applicability.
• Adapted conditional likelihood methods for random intercept models with canonical links.

Main Results:

• The proposed methods are applicable to various outcome-related sampling designs.
• Demonstrated the utility of these methods using data from a study on attention deficit hyperactivity disorder.
• The techniques offer improved estimation efficiency for longitudinal data.

Conclusions:

• The developed statistical methods effectively analyze longitudinal data from outcome-related samples.
• These approaches provide valuable tools for researchers dealing with rare or costly outcomes.
• The findings have implications for improving the cost-effectiveness of longitudinal health studies.