Laplace regression with censored data.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Accurate estimation of conditional quantiles is crucial for understanding outcome variables.
• Random censoring, potentially dependent on covariates, complicates standard regression analyses.
• Existing methods for quantile estimation under censoring may suffer from convergence issues and longer computation times.

Purpose of the Study:

• To propose and evaluate a novel regression model using an asymmetric Laplace distribution for estimating conditional quantiles.
• To address challenges posed by random censoring, including covariate-dependent censoring.
• To compare the performance of the proposed Laplace estimator against alternative methods.

Main Methods:

• Development of a regression model incorporating an asymmetric Laplace distribution for error terms.
• Estimation of regression coefficients via maximization of a non-differentiable likelihood function.
• Simulation studies to assess the performance of the proposed method.

Main Results:

• The Laplace estimator demonstrated correct coverage in simulation scenarios.
• The proposed method exhibited significantly shorter computation times compared to alternatives.
• Some alternative methods encountered convergence failures during the simulation study.

Conclusions:

• The asymmetric Laplace regression model provides a robust and efficient approach for quantile estimation with randomly censored data.
• This method is particularly advantageous when dealing with covariate-dependent censoring.
• The approach shows promise for applications in survival analysis, such as in small cell lung cancer patient data.