Regression analysis of current status data with latent variables

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Current status data are prevalent across diverse fields like epidemiology, finance, and sociology.
• Analyzing such data, especially with unobserved latent variables, presents unique statistical challenges.
• Existing methods may not fully capture the complexities introduced by latent structures in current status data.

Purpose of the Study:

• To develop a robust regression analysis framework for current status data incorporating latent variables.
• To model latent variables using multiple surrogates via a factor analytic approach.
• To investigate covariate effects on event hazards using an additive hazard model.

Main Methods:

• A novel estimation procedure combining the expectation-maximization algorithm with correlated estimating equations.
• Development of a borrow-strength estimation technique tailored for current status data with latent variables.
• Theoretical establishment of the consistency and asymptotic normality of the proposed estimators.

Main Results:

• The proposed method effectively handles current status data with latent variables.
• The borrow-strength estimation procedure demonstrates favorable performance in simulations.
• Consistency and asymptotic normality of estimators provide theoretical validation for the approach.

Conclusions:

• The developed methodology offers a powerful tool for analyzing complex current status data in various scientific domains.
• The approach is validated through simulation studies and demonstrated on a real-world chronic kidney disease dataset.
• This work contributes to advancing statistical methods for handling latent variables in survival analysis.