Semiparametric likelihood inference for left-truncated and right-censored data

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Survival data often presents complexities like left-truncation and right-censoring.
• Accurate estimation requires accounting for the truncation time distribution, which is often unknown.
• Existing models like the Vardis multiplicative censoring model offer a foundation but require extension.

Purpose of the Study:

• To develop a novel estimation procedure for survival time distributions with left-truncated and right-censored data.
• To extend the Vardis multiplicative censoring model for this data type.
• To introduce a formal test for the truncation time distribution and assess stationarity assumptions.

Main Methods:

• Utilizing a generalized multiplicative censoring model framework.
• Deriving an Expectation-Maximization (EM) algorithm for parameter estimation.
• Constructing a semiparametric likelihood ratio test for the truncation time distribution.

Main Results:

• The proposed method provides a robust estimation procedure for complex survival data.
• A formal test successfully evaluates the truncation time distribution, including stationarity.
• Asymptotic properties of the estimator are theoretically established and validated through simulations.

Conclusions:

• The developed Expectation-Maximization algorithm and likelihood ratio test offer significant advancements in survival data analysis.
• The methods are applicable to real-world datasets, demonstrating their practical utility.
• Analysis of the Canadian Study of Health and Aging and Channing House data highlights the importance of assessing stationarity assumptions.