Mixed model analysis of censored longitudinal data with flexible random-effects density

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Statistical Modeling

Background:

• Mixed models are standard for longitudinal data.
• Censored responses, common in biological assays, complicate standard analysis.
• The impact of misspecifying random-effects distributions in censored models is not well understood.

Purpose of the Study:

• To investigate the consequences of misspecifying random-effects distributions in mixed models for censored longitudinal data.
• To develop and evaluate a more flexible mixed model framework for censored data.
• To assess the performance of the proposed method using simulations and real-world data.

Main Methods:

• Developed a mixed model framework incorporating a flexible seminonparametric density for random effects.
• Utilized maximum likelihood estimation within the SAS procedure NLMIXED.
• Conducted simulations to compare the proposed method with standard Gaussian random-effects models.

Main Results:

• Maximum likelihood estimators can be inconsistent when random-effects distributions are misspecified.
• The proposed seminonparametric approach demonstrated reduced bias and increased efficiency compared to Gaussian assumptions.
• The benefits were most pronounced when true random-effects distributions deviated from normality and noncensored data was sparse.

Conclusions:

• Flexible seminonparametric random-effects distributions offer significant advantages for censored longitudinal data analysis.
• The developed methods provide a robust alternative to standard Gaussian assumptions, improving statistical inference.
• The approach is applicable to various fields, including virology studies like the hepatitis C virus example.