Joint regression analysis of clustered current status data with latent variables

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Clustered current status data are common in survival studies.
• Unobserved factors (latent variables) and cluster size can influence outcomes.
• Existing methods may not fully account for these complexities.

Purpose of the Study:

• To propose a joint modeling method for analyzing clustered current status data.
• To incorporate latent variables and potentially informative cluster sizes.
• To provide a robust statistical framework for survival data with complex dependencies.

Main Methods:

• A joint model combining factor analysis for latent variables and an additive hazards frailty model.
• Utilizing the expectation-maximization algorithm and weighted estimating equations for parameter estimation.
• Establishing theoretical properties including consistency and asymptotic normality of estimators.

Main Results:

• The proposed method effectively models latent variables using surrogate markers.
• It accounts for intra-cluster correlations and informative cluster sizes.
• Simulation studies demonstrate good finite-sample performance.

Conclusions:

• The developed joint modeling approach offers a powerful tool for clustered current status data.
• It is applicable to various fields, including toxicology and health research.
• The method provides reliable estimates for covariate effects in complex survival data.