A Corrected Score Approach for Proportional Hazards Model With Error-Contaminated Covariates Subject to Detection...

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Covariates in proportional hazards models can have measurement error and detection limits.
• Existing methods often address only one issue, leading to bias and incorrect inference when both are present.
• Limited research tackles both measurement error and detection limits simultaneously, often relying on restrictive assumptions.

Purpose of the Study:

• To develop a novel statistical approach for survival analysis that accounts for both covariate measurement error and detection limits.
• To overcome the limitations of existing likelihood-based methods that require strong distributional and independence assumptions.
• To provide a computationally simpler and more robust estimation method.

Main Methods:

• A corrected score approach is proposed to address simultaneous measurement error and detection limits.
• The method alleviates stringent distributional assumptions on true covariates and independence assumptions on censoring time.
• The approach is applicable to replicate or instrumental data and can be extended to other models.

Main Results:

• The proposed corrected score estimator is demonstrated to be consistent and asymptotically normal.
• Simulation studies assess the finite sample performance of the new estimator.
• The method is illustrated using data from an AIDS clinical trial.

Conclusions:

• The novel corrected score approach effectively handles both measurement error and detection limits in survival analysis.
• This method offers a more flexible and computationally efficient alternative to existing techniques.
• The approach has broad applicability and potential for extension to more complex statistical scenarios.