Right-censored partially linear regression model with error in variables: application with carotid endarterectomy...

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Partially linear regression models are widely used but face challenges with right-censored data and covariates with measurement error.
• Accurate estimation requires addressing both censorship and measurement error simultaneously.
• Existing methods may not adequately handle the complexities of these combined issues.

Purpose of the Study:

• To develop and evaluate novel semiparametric estimators for partially linear regression models with right-censored responses and covariates subject to measurement error.
• To compare the performance of different smoothing techniques in the presence of both censorship and measurement error.
• To provide a robust estimation method for real-world applications involving complex data structures.

Main Methods:

• Proposed three modified semiparametric estimators utilizing local polynomial regression, kernel smoothing, and B-spline smoothing.
• Employed a kernel deconvolution approach to address the measurement error problem.
• Utilized synthetic data transformation to incorporate the effect of censorship into the estimation procedure.

Main Results:

• A detailed Monte Carlo simulation study was conducted to compare the performance of the proposed estimators.
• The deconvoluted local polynomial method yielded more qualified estimates compared to the other two methods.
• The methods were validated using real-world Carotid endarterectomy data.

Conclusions:

• The proposed semiparametric estimators effectively handle partially linear regression models with both right-censored responses and covariates with measurement error.
• The deconvoluted local polynomial approach offers a superior estimation strategy for such models.
• The findings have implications for statistical modeling in fields where censored data and measurement error are prevalent.