Corrected likelihood for proportional hazards measurement error model and its application

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Accurate covariate data is crucial for proportional hazards models.
• Measurement errors in covariates can lead to biased statistical estimates.
• Existing methods may not adequately address covariate measurement error in survival data.

Purpose of the Study:

• To investigate the impact of measurement error in covariates on proportional hazards models.
• To evaluate a proposed correction method for partial likelihood in the presence of measurement error.
• To assess the utility of this correction method in practical survival analysis applications.

Main Methods:

• Utilized Monte Carlo simulations to illustrate bias from using surrogate covariates.
• Analyzed a proposed correction to the partial likelihood function.
• Reviewed and discussed alternative measurement error correction techniques for censored survival models.

Main Results:

• Maximum likelihood estimates using surrogate covariates are asymptotically biased.
• The proposed partial likelihood correction method demonstrates utility when the effective magnitude of measurement error is small.
• Simulation results highlight the potential for misleading conclusions without correction.

Conclusions:

• Covariate measurement error significantly impacts proportional hazards model validity.
• The corrected partial likelihood approach offers a valuable solution for survival analysis with measurement error, particularly when error is minimal.
• Careful consideration of measurement error is essential for reliable survival data analysis.