Estimating equations with incomplete categorical covariates in the Cox model

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Incomplete covariate data is frequent in survival time studies.
• The Expectation-Maximization (EM) algorithm is useful for parameter estimation with full likelihoods.
• Estimating parameters for Cox's proportional hazards model with missing data is challenging.

Purpose of the Study:

• To propose a set of estimating equations for Cox's proportional hazards model with missing covariate values.
• To develop an algorithm similar to the EM algorithm for solving these equations.
• To offer a computationally feasible method for parameter estimation using Monte Carlo techniques.

Main Methods:

• Developed estimating equations for Cox's proportional hazards model with incomplete covariate data.
• Adapted the EM algorithm to solve the proposed estimating equations.
• Utilized Monte Carlo methods to overcome computational burdens.
• Derived asymptotic variances for parameter estimates.

Main Results:

• The proposed method effectively estimates parameters in Cox's proportional hazards model with missing covariates.
• Monte Carlo methods provide a practical solution for computational challenges.
• Asymptotic variances of the parameter estimates were successfully derived.
• Demonstrated the method's application using a clinical trial example with missing covariate data.

Conclusions:

• The proposed method offers a robust approach to handling missing covariate data in survival analysis.
• Monte Carlo-based EM algorithm provides an efficient way to estimate parameters.
• The technique is applicable to real-world scenarios, such as clinical trials.