on June 12, 2000
Reliability and survival analysis both deal with time to failure data. Much of the methodology is essentially the same. The term reliability is generally used to apply to hardware or software whereas survival analysis is a term for biological systems such as animals or humans. This book includes the standard nonparametric and parametric methods for estimating reliability functions and parameters. It includes system reliability and repairable systems and deals with recent developments with repairable systems including Nelson's mean cumulative function. A couple of years ago I asked Wayne Nelson if and when he might revise his popular text "Applied Life Data Analysis". He said he did not plan to do it because Meeker and Escobar had just finished a work that would be as good as any revision he might want to produce. Other topics include failure time regression models including the popular Cox proportional hazards model and accelerated life test models. It also includes modern topics such as bootstrap confidence intervals (both semi-parametric and nonparametric) for reliability parameters. The book is comprehensive and up-to-date. It also includes discussion of Bayesian methods. Some case studies are also included. The only topics it misses are reliability growth and warranty and service contracts. These topics are covered in the recent book by Blischke and Murthy "Reliability Modeling, Prediciton, and Optimization" also published by John Wiley and Sons, Inc.
Numerical examples are done using the SPlus software from MathSoft. An ftp site is available to download data sets to use with SPlus.
on August 16, 1998
Reliability data isn't amenable to treatment with the ubiquitous normal distribution, a fact which should catch the attention of any practicing engineer with only that bullet in his gun. Reliability data has other distinguishing features. The data are usually censored, which means the exact failure times are not known so the observations can only provide bounds on the actual failure times. Inferences and predictions usually require extrapolations, making engineering and physics-based modeling an important adjunct to statistical methods. Whereas many *statistical* problems focus on parameter estimation (e.g.: mean, standard deviation), these are not of primary interest to engineers who need specific measures of product reliability (e.g.: failure probabilities, life distribution quintiles, failure rates).
The chapter headings provide an overview of the book:
1) Reliability Concepts and Reliability Data 2) Models, Censoring, and Likelihood for Failure-Time Data 3) Nonparametric Estimation 4) Location-Scale-Based Parametric Distributions 5) Other Parametric Distributions 6) Probability Plotting 7) Parametric Likelihood Fitting Concepts: Exponential Distribution 8) Maximum Likelihood for Log-Location-Scale Distributions 9) Bootstrap Confidence Intervals 10) Planning Life Tests 11) Parametric Maximum Likelihood: Other Models 12) Prediction of Future Quantiles 13) Degradation Data, Models, and Data Analysis 14) Introduction to the Use of Bayesian Methods for Reliability Data 15) System Reliability Concepts and Methods 16) Analysis of Reparable System and Other Recurrence Data 17) Failure-Time Regression Analysis 18) Accelerated Life Tests 21) Accelerated Degradation Tests 22) Case Studies and Further Applications Appendix A - Notation and Acronyms Appendix B - Some Results from Statistical Theory
This book is written for practitioners - engineers and statisticians - yet does not presume an undergraduate degree in statistics. More involved statistical ideas (Bayesian thought, censored observations, bootstrapping, et cetera) are all described to the user with the assumption that they have had little prior exposure. The book's concepts are presented in an unstuffy and intuitive manner. For example, for Meeker and Escobar likelihood is simply "the probability of the data," making a maximum likelihood estimator one which maximizes the probability that the experiment turned out the way it did. (Contrast this to the hushed tones in many "engineering statistics" texts which suggest that Likelihood is a profound concept beyond the (limited) capacity of the engineer and best left to the trained statisticians.) The wholesome, unpretentious, and practical approach taken by Meeker and Escobar is quite pleasing to this reviewer, a professional engineer whose formal statistical education began later in life.
The book should be interesting to statisticians too. It can be used as a two-semester graduate statistics course, a one-semester course for engineers and statisticians, or as the basis for workshops and short courses on selected topics for industry practitioners. Each chapter is suffused with examples using real data and ends with thought-provoking exercises. While this is a practical book, it does not neglect statistical theory (after all, the authors are well-known academic statisticians) - although it is interesting to note that for censored observations there may be no *exact* theory for statistical inference. While the book's emphasis is more on results than on theoretical proofs, I think the practicing statistician will be quite pleased with the book's balance.
Not only are its 680 pages chock-full of ideas, information, and techniques, _Statistical Methods for Reliability Data_ is a noteworthy paradigm for technical exposition: Even before each chapter's introduction, there is a brief statement of chapter objectives, followed by an overview which places the chapter in perspective, stating for example, that the material is a prerequisite for this or that future topic, or conditions under which it could be omitted, or why its is useful. This makes it easy for a practitioner to find his way around the text.
In summary: Buy this book. If competitive advantage through reliable products is central to your company's future, then Meeker and Escobar, _Statistical Methods for Reliability Data_ can help you reach your objectives.
on September 20, 2001
The purpose of this book was supposed to serve very broad groups of people: students, statisticians and engineers. Unfortunately, I found this book not quite suitable in engineering practice.
From practical point of view, when dealing with reliability estimations, one has to connect mathematical theory with real-life data. It appears that to accomplish this task it is necessary to understand some basic statistical ideas, plus specifics of the subject under consideration. Sometimes common sense knowledge can come in handy. Strangely enough but many fundamental principles are in fact surprisingly simple, elegant and thus beautiful. What is missing in the book is the lack of clear explanations of fundamental statistical concepts that certainly can be presented in a complicated form but in reality they are not. On the other side, the book could serve as a solid textbook to students, statisticians and mathematicians.