|
Activity Number:
|
383
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #304104 |
|
Title:
|
Inference with Censored Degradation Data
|
|
Author(s):
|
Yang Yang*+ and Vijay Nair
|
|
Companies:
|
University of Michigan and University of Michigan
|
|
Address:
|
439 West Hall , Ann Arbor, MI, 48109-1107,
|
|
Keywords:
|
degradation data ; censoring ; EM algorithm ; autoregressive models
|
|
Abstract:
|
Degradation data are commonly measured over time to investigate the physical deterioration of products. We have been involved in a project to analyze distress indices which are recorded to evaluate the condition of road pavements. The goal of the project was to determine the effect of various pavement design parameters and predict performance and the need for repairs over time. The database was very incomplete with missing observations as well as left and right censoring. In this paper, we discuss methods for analyzing this type of censored degradation data and for making predictions at the individual device level. Unlike the estimation of failure time distributions, nonparametric estimation appears difficult, so we focus on Gaussian models with a time series structure. Maximum likelihood estimation, use of EM algorithm, and connections to GEE models are discussed.
|
