Robust best linear weighted estimator with missing covariates in survival analysis

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Missing covariate data can lead to biased estimates and reduced statistical power in regression analyses.
• Inverse probability weighting (IPW) is a common method for handling missing covariates but can be less efficient than likelihood-based methods.
• Robustness against model misspecification is crucial in real-world data analysis.

Purpose of the Study:

• To propose a novel robust best linear weighted estimator for Cox regression with missing covariates.
• To enhance the efficiency of estimators in the presence of missing covariate data.
• To provide a statistically sound method for analyzing survival data with incomplete covariate information.

Main Methods:

• Developed a robust best linear weighted estimator by projecting the IPW estimator onto an orthogonal complement.
• Utilized a working regression model of observed data to leverage associations between survival outcomes and available covariates.
• Derived the asymptotic distribution of the proposed estimator.
• Conducted extensive simulation studies to evaluate finite sample performance.

Main Results:

• The proposed robust best linear weighted estimator demonstrates improved efficiency compared to standard IPW estimators.
• The estimator maintains robustness against potential model misspecification.
• Simulation studies confirm the favorable finite sample performance of the new method.

Conclusions:

• The robust best linear weighted estimator offers a valuable advancement for Cox regression with missing covariates.
• This method provides a more efficient and reliable approach to survival data analysis when covariate data is incomplete.
• The approach was successfully applied to a colorectal cancer dataset, demonstrating its practical utility.