GENERALIZED M-ESTIMATION FOR NONLINEAR CENSORED REGRESSION

Siyang WANG,Tao HU,Hengjian CUI
2014-01-01
Abstract:Accelerated failure time model is an important regression tool to study association between failure time and co-variates.In standard accelerated lifetime model,log lifetime is a linear function of co-variable vector.This model is now extended to admit general nonlinear functional relationships.We propose a robust weighted generalized M-estimation for nonlinear regression with right-censored data using the Kaplan-Meier weights in the generalized M type obj ective function to estimate regression coefficients and scale parameter simultaneously.Asymptotic properties including the root-n consistency and asymptotic normality are established for the resulting estimator under suitable conditions. Simulations suggest that the proposed estimator has good finite sample performance in comparison with existing robust estimators.
What problem does this paper attempt to address?