Preface to the Second Edition

The first edition of Statistical Methods for Reliability Data (SMRD1) was published 23 years ago. We believe SMRD1 successfully met its goal of providing a comprehensive overview of statistical methods for reliability data analysis and test planning for practitioners and statisticians and we have received much positive feedback. Despite, and perhaps because of this, there were compelling reasons for a second edition (SRMD2). We (both the profession and the authors) have learned much since 1998. In our experiences consulting on reliability applications and in presenting on the order of one hundred short courses to both statisticians and reliability engineers in industry and government, we have gained a solid appreciation for what are the most important topics in the field of statistical methods for reliability, suggesting a slight change in focus for SMRD2. In teaching our respective university courses, we have discovered some improved methods for presenting some topics. These experiences helped us to develop our plan for SMRD2. For the SMRD2 project, Bill and Luis have been most fortunate to have Jave Pascual join them as a co‐author.

Goals for SMRD2

Our goals for the SMRD2 were to:

• Update and improve or expand on various previously presented statistical methods for reliability data, using statistical and computational methods that have been developed or become readily available since SMRD1 was published.
• Improve the organization of the material to make it possible to cover more and the most important topics in a one‐semester course.
• Completely rewrite chapters where there have been important developments or changes in what are considered to be best practices in the analysis of reliability data.
• Provide a more extensive treatment of the use and application of Bayesian methods in reliability data analysis.
• Provide, via technical appendices, additional justification and theory underlying the statistical methods presented in this book.
• Provide a webpage that gives up‐to‐date information about available software for doing reliability data analysis, supplementary information such as presentation slides, additional data sets, and exercises, as well as other important up‐to‐date information developed or coming to our attention after publication.

What Has Not Changed

In SMRD2, we have paid special care to retain the appealing features of SMRD1. Specifically, the special features of the book, listed in the preface of SMRD1 are still intact.

Important Changes in SMRD2

Important changes in SMRD2 include the following:

• Although SMRD1 has been popular with both engineers and statisticians, in the preparation of SMRD2 there has been a concerted effort to look for ways to improve the presentation and usability of the book. Means for doing this have included additional words of explanation, additional examples using simulation to provide insights, moving some technical details to appendices, and omitting topics that, while of some technical interest, have had little or no value in practical applications.
• We have added a section on the important topic of the distribution of remaining life to the chapter Models, Censoring, and Likelihood for Failure‐Time Data (Chapter 2).
• We now include a section describing the important Fréchet distribution in the chapter Some Parametric Distributions Used in Reliability Applications (Chapter 4).
• SMRD1 Chapter 5 (Other Parametric Distributions) has been eliminated, with the important and useful material being moved to either Chapter 4 or the chapter Special Parametric Models (Chapter 11).
• The chapter Parametric Bootstrap and Other Simulation‐Based Confidence Interval Methods (Chapter 9), has been completely rewritten to reflect many new developments since SMRD1 was published, including the use of the fractional‐random‐weight bootstrap and generalized pivotal quantities.
• The chapter An Introduction to Bayesian Statistical Methods for Reliability (Chapter 10) replaces, updates, and expands SMRD1 Chapter 14. The chapter has been completely rewritten with a more modern slant on prior distribution specification and computational methods, with several illustrations of Bayesian applications. Then, in most subsequent chapters, Bayesian methods are integrated into the development and presentation of many statistical analyses where some prior information is available (and ignoring it would be wrong) or where there is other strong motivation to use Bayesian methods.
• SMRD1 Chapter 10 has been completely rewritten to focus on Planning Life Tests for Estimation (now Chapter 13) and to improve clarity and usability of the material.
• The material in the chapter Planning Reliability Demonstration Tests (Chapter 14) is mostly new, where the SMRD1 material (previously in a small part of SMRD1 Chapter 10) has been completely rewritten and generalized to allow planning demonstration tests for any log‐location‐scale distribution and tests that allow failures to occur (and still pass the test), providing demonstration tests that have much improved probability of successful demonstration (i.e., power).
• The chapter Prediction of Failure Times and the Number of Future Field Failures (Chapter 15) has also been completely rewritten to reflect simpler and more direct methods to obtain prediction intervals in reliability applications. These include the use of predictive distributions for both non‐Bayesian and Bayesian prediction. Also, we now put more focus on the important applications of predicting the number of field failures and warranty returns.
• Instead of one combined chapter on system reliability and the analysis of competing risks (SMRD1 Chapter 15), these topics are now covered more extensively in two separate chapters: System Reliability Concepts and Methods (Chapter 5) and Analysis of Data with More than One Failure Mode (Chapter 16), respectively.
• The material on regression analysis of failure‐time data and accelerated life testing has been reorganized with many improvements and new examples, now in Chapters 17, 18, and 19.
• We have added a completely new chapter, Degradation Modeling and Destructive Degradation Data Analysis (Chapter 20), to describe and illustrate the use of these important statistical methods.
• To save space and improve organization, the two SMRD1 chapters on degradation modeling and analysis (SMRD1 Chapters 13 and 21) have been combined and completely rewritten to form Repeated‐Measures Degradation Modeling and Analysis (Chapter 21). To make inferences from repeated‐measures degradation data, we have replaced the previously used maximum likelihood and bootstrap methods for nonlinear mixed‐effects models with the more versatile Bayesian hierarchical models and inference methods.
• Almost all of the figures in SMRD2 have been redrawn, both to improve quality and to introduce color (for the electronic versions of SMRD2). We have, however, continued to design our graphics so that having color is not necessary.
• Numerous new references have been added or updated and the Bibliographic Notes and Related Topics sections at the end of each chapter have been expanded and reorganized to make it easier to find references and additional information for particular topics.
• Many new data sets and examples have been added throughout SMRD2. We also have added many new applications to illustrate the use of the methods that we present. As in SMRD1, all of these applications are based on real data. In some of the data sets, however, we have changed the names of the variables and/or the scale of the data to protect sensitive information. The data in Section 23.3 had to be simulated, but they reflect the interesting statistical aspects of the real applications.
• Many new exercises have been added at the end of the chapters.
• Some tables containing small reliability data sets remain in the chapters and in Appendix D. These data sets and all new data sets, used in examples and exercises, are available as csv files from the SRMD2 webpage.

What Was Dropped

Some material from SMRD1 Chapter 5 (Other Parametric Distributions) and Chapter 20 (Planning Accelerated Life Tests) has been dropped. As mentioned earlier, the useful material from Chapter 5 has been integrated into either Chapter 4 or Chapter 11. The important ideas behind planning accelerated life tests are illustrated and briefly described in some of the accelerated testing examples in Chapters 18 and 19 with a summary of the key points in Section 19.4.3. To save space for more important material, we have also dropped a few tables of lengthy data sets. These data sets (and many others) are, however, available on SMRD2's webpage.

Overview and Paths through SMRD2

There are many paths that readers and instructors might take through this book. Chapters 1–8 cover single distribution models without any explanatory variables. This is basic material that will be of interest to almost all readers and should be read in sequence. It is possible to do only a light...