Induced Smoothing for Rank-Based Accelerated Failure Time Models with General Weight Functions

• Activity Number: 64
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Jun Yan*+ and Sy Han Chiou and Sangwook Kang
• Companies: University of Connecticut and University of Connecticut and University of Connecticut

Abstract

The induced smoothing approach for rank-based accelerated failure time (AFT) models is natural with Gehan's weight. When other weights are used, the induced smoothing approach does not in general provide smooth estimating equations that are easy to evaluate. By using the estimator from Gehan's weight as initial value, we propose an iterative induced smoothing procedure that works for any general weight function. The resulting estimator has the same asymptotic properties as the un-smooth rank-based estimator with general weights. The variance of the estimator is estimated by an efficient resampling approach that does not require solving estimating equations repeatedly. Finite sample performance of the estimator is shown to be good and competitive with other estimators in a numerical study. The implementation is available in an R package aftgee.