Premium
Refined Rank Regression Method with Censors
Author(s) -
Wang Wendai
Publication year - 2004
Publication title -
quality and reliability engineering international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.913
H-Index - 62
eISSN - 1099-1638
pISSN - 0748-8017
DOI - 10.1002/qre.575
Subject(s) - rank (graph theory) , computer science , convergence (economics) , suspension (topology) , regression , regression analysis , parametric statistics , reliability (semiconductor) , mathematics , statistics , combinatorics , homotopy , pure mathematics , economics , economic growth , power (physics) , physics , quantum mechanics
Reliability engineers often face failure data with suspensions. The rank regression method with an approach introduced by Johnson has been commonly used to handle data with suspensions in engineering practice and commercial software. However, the Johnson method makes partial use of suspension information only—the positions of suspensions, not the exact times to suspensions. A new approach for rank regression with censored data is proposed in this paper, which makes full use of suspension information. Taking advantage of the parametric approach, the refined rank regression obtains the ‘exact’ mean order number for each failure point in the sample. With the ‘exact’ mean order number, the proposed method gives the ‘best’ fit to sample data for an assumed times‐to‐failure distribution. This refined rank regression is simple to implement and appears to have good statistical and convergence properties. An example is provided to illustrate the proposed method. Copyright © 2004 John Wiley & Sons, Ltd.