ASA-SIAM Series on Statistics and Applied Probability – serie

American industry is becoming more aware of the importance of applying statistical methods to improve its competitive edge in the world market. Examples of real industrial applications can serve as a major motivator for industries that want to increase their use of statistical methods.This book contains a broad selection of case studies written by professionals in the semiconductor industry that illustrate the use of statistical methods to improve manufacturing processes. These case studies offer engineers, scientists, technicians, and managers numerous examples of best-in-class practices by their peers. Because of the universal nature of statistical applications, the methods described here can be applied to a wide range of industries, including the chemical, biotechnology, automotive, steel, plastics, textile, and food industries. Many industries already benefit from the use of statistical methods, although the semiconductor industry is considered both a leader in and a model for the wide application and effective use of statistics.Specific case studies address the following statistical methods: gauge studies, passive data collection (sources of variation studies), design of experiments, statistical process control, and equipment reliability.Readers familiar with the statistical methodologies that comprise the Six Sigma® tool box will find a wealth of applications. Czitrom has written an introduction to each statistical method, which, along with a glossary, gives basic definitions of frequently occurring statistical terms and suggestions for further reading. The case studies, which can be used in industrial training as well as in academia, are an extremely useful classroom supplement and will remain a rich source of used and useful approaches to real industrial problems for years to come.

Engineering reliability concerns failure data analysis, the economics of maintenance policies, and system reliability. This textbook develops the use of probability and statistics in engineering reliability and maintenance problems. The author uses probability models in the analysis of failure data, decision relative to planned maintenance, and prediction relative to preliminary design.Some of the outstanding features include:The analysis of failure data for both continuous and discrete probability from a finite population perspective.Probability models derived from engineering considerations.An introduction to influence diagrams and decision making.Use of the operational bayesian approach.The approach is fresh and interesting; it is motivated from problems in engineering and physical sciences and uses examples to illustrate the methodology. These examples, along with the use of real failure time data, will help the reader apply the techniques to real industrial situations.This upper undergraduate or first-year graduate level book is largely self-contained relative to the background material required. Each chapter contains exercises, notes, references, and notation. Spreadsheets are used extensively in the exercises, and are also useful for the graphical analysis of failure data contained in the book.A complimentary instructor's manual is available upon adoption.

Statisticians know that the clean data sets that appear in textbook problems have little to do with real-life industry data. To better prepare their students for all types of statistical careers, academic statisticians now strive to use data sets from real-life statistical problems. This book contains 20 case studies that use actual data sets that have not been simplified for classroom use. Each case study is a collaboration between statisticians from academe and from business, industry, or government.This book is the result of a collaborative workshop of statisticians focusing on academic?industrial partnerships. The cases come from a wide variety of application areas, including biology/environment, medical and health care, pharmaceutical, marketing and survey research, and manufacturing.

Statisticians know that the clean data sets that appear in textbook problems have little to do with real-life industry data. To better prepare their students for all types of statistical careers, academic statisticians now strive to use data sets from real-life statistical problems. This book contains 20 case studies that use actual data sets that have not been simplified for classroom use. Each case study is a collaboration between statisticians from academe and from business, industry, or government.This book is the result of a collaborative workshop of statisticians focusing on academic?industrial partnerships. The cases come from a wide variety of application areas, including biology/environment, medical and health care, pharmaceutical, marketing and survey research, and manufacturing.

Matrix analytic methods are popular as modeling tools because they give one the ability to construct and analyze a wide class of queuing models in a unified and algorithmically tractable way. The authors present the basic mathematical ideas and algorithms of the matrix analytic theory in a readable, up-to-date, and comprehensive manner. In the current literature, a mixed bag of techniques is used—some probabilistic, some from linear algebra, and some from transform methods. Here, many new proofs that emphasize the unity of the matrix analytic approach are included.The authors begin by describing several examples of quasi-birth-and-death (QBD) processes. These examples give the reader an idea of the variety of models which are hidden by the general block notation as well as reinforce some of the terminology and notation used throughout the text. These same examples are used as illustrations later. The second part of the book deals with phase-type distributions and related-point processes, which provide a versatile set of tractable models for applied probability. Part three reviews birth-and-death processes, and points out that the arguments for these processes carry over to more general processes in a parallel manner and are based on Markov renewal theory.Part four covers material where algorithmic and probabilistic reasoning are most intimately connected. In three steps, the authors take you from one of the simplest iterative procedures to the fastest, relating the successive approximations to the dynamic behavior of the stochastic process itself. The final part goes beyond simple QBDs with a sequence of short chapters where the authors discuss various extensions to the analyzed processes. Their intention is to show that the fundamental ideas extend beyond simple homogeneous QBD.

Expert judgment is invaluable for assessing products, systems, and situations for which measurements or test results are sparse or nonexistent. This book takes the reader step by step through the techniques of eliciting and analyzing expert judgment, with special attention given to helping the reader develop elicitation methods and tools adaptable to a variety of unique situations and work areas.The analysis procedures presented in the book may require a basic understanding of statistics and probabilities, but the authors have provided detailed explanations of the techniques used and have taken special care to define all statistical jargon.Originally published in 1991, this book is designed so that those familiar with the use of expert judgment can quickly find the material appropriate for their advanced background. Newcomers will be eased into the topic by special chapters explaining what expert judgment is and how it can be used, an introduction to statistical techniques and notation, definitions for statistical terms, and a separate glossary of expert judgment terms.People in industry, government, and academia are aware that they must capture their expertise, as well as access and update it in electronic repositories (knowledge systems), if they are to preserve their chief assets and competitive edge. This book provides a shortcut to creating these knowledge systems.The innovative methodology PREDICT (Performance and Reliability Evaluation from Diverse Information, Combination, and Tracking) is used to predict the performance of a complex system undergoing change. This book provides PREDICT users with guidance on how to elicit and analyze expert judgment, including the associated uncertainty. Expert judgment is an important source of information when test data is not available.

This applied, self-contained text provides detailed coverage of the practical aspects of multivariate statistical process control (MVSPC) based on the application of Hotelling's T2 statistic. MVSPC is the application of multivariate statistical techniques to improve the quality and productivity of an industrial process. The authors, leading researchers in this area who have developed major software for this type of charting procedure, provide valuable insight into the T2 statistic. Intentionally including only a minimal amount of theory, they lead readers through the construction and monitoring phases of the T2 control statistic using numerous industrial examples taken primarily from the chemical and power industries. These examples are applied to the construction of historical data sets to serve as a point of reference for the control procedure and are also applied to the monitoring phase, where emphasis is placed on signal location and interpretation in terms of the process variables.Specifically devoted to the T2 methodology, this is the only book available that concisely and thoroughly presents such topics as how to construct a historical data set; how to check the necessary assumptions used with this procedure; how to chart the T2 statistic; how to interpret its signals; how to use the chart in the presence of autocorrelated data; and how to apply the procedure to batch processes. The book comes with a CD-ROM containing a 90-day demonstration version of the QualStat™ multivariate SPC software specifically designed for the application of T2 control procedures. The CD-ROM is compatible with Windows® 95, Windows® 98, Windows® Me Millennium Edition, and Windows NT® operating systems.

Survival data consist of a single event for each population unit, namely, end of life, which is modeled with a life distribution. In contrast, many applications involve repeated-events data, where a unit may accumulate any number of events over time. Examples include the number and cost of repairs of products, the number and treatment costs of recurrent disease episodes in patients, and the number of childbirths to statisticians. This applied book provides practitioners with basic nonparametric methods for such data, particularly the plot of the estimate of the population mean cumulative function (MCF), which yields most of the information sought.Recurrent Events Data Analysis for Product Repairs, Disease Recurrences, and Other Applications is the first book to present a simple, unified theory that includes data on costs or other ""values"" of discrete events, not just the number of events. It surveys computer programs that calculate and plot the MCF estimate with confidence limits, shows their output, and explains how to interpret such plots. Many such calculations can be easily done with a pocket calculator or spreadsheet program. Also, the book introduces basic Poisson and Cox regression models and parametric models, including homogeneous and nonhomogeneous Poisson processes and renewal processes.

Probabilists and fuzzy enthusiasts tend to disagree about which philosophy is best and they rarely work together. As a result, textbooks usually suggest only one of these methods for problem solving, but not both. This book, with contributions from 15 experts in probability and fuzzy logic, is an exception. The contributing authors, investigators from both fields, have combined their talents to provide a practical guide showing that both fuzzy logic and probability have their place in the world of problem solving. They work together with mutual benefit for both disciplines, providing scientists and engineers with examples of and insight into the best tool for solving problems involving uncertainty.Fuzzy Logic and Probability Applications makes an honest effort to show both the shortcomings and benefits of each technique, and even demonstrates useful combinations of the two. It provides clear descriptions of both fuzzy logic and probability, as well as the theoretical background, examples, and applications from both fields, making it a useful hands-on workbook for members of both camps. It contains enough theory and references to fundamental work to provide firm ground for both engineers and scientists at the undergraduate level and above. Readers should have a familiarity with mathematics through calculus.Use of this book is not restricted to a specific course or application. It can be used in teaching probability, fuzzy logic, general problem solving, or in any course in which probability and fuzzy logic are not normally taught together. It has applications in control theory and artificial intelligence, knowledge acquisition/management, and risk/reliability analysis.

Adaptive statistical tests, developed over the last 30 years, are often more powerful than traditional tests of significance, but have not been widely used. To date, discussions of adaptive statistical methods have been scattered across the literature and generally do not include the computer programs necessary to make these adaptive methods a practical alternative to traditional statistical methods. Until recently, there has also not been a general approach to tests of significance and confidence intervals that could easily be applied in practice.Modern adaptive methods are more general than earlier methods and sufficient software has been developed to make adaptive tests easy to use for many real-world problems. Applied Adaptive Statistical Methods introduces many of the practical adaptive statistical methods developed over the last 10 years and provides a comprehensive approach to tests of significance and confidence intervals. It shows how to make confidence intervals shorter and how to make tests of significance more powerful by using the data itself to select the most appropriate procedure.Adaptive tests can be used for testing the slope in a simple regression, testing several slopes in a multiple linear regression, and for the analysis of covariance. The increased power is achieved without compromising the validity of the test, by using adaptive methods of weighting observations and by using permutation techniques. An adaptive approach can also be taken to construct confidence intervals and to estimate the parameters in a linear model. Adaptive confidence intervals are often narrower than those obtained from traditional methods and maintain the same coverage probabilities.Numerous applied examples from the areas of biostatistics, health sciences, the pharmaceutical industry, agricultural sciences, education, and environmental science are included. The SAS macros discussed in the text are provided in the Appendix and can also be downloaded from the author’s website.

Bayesian Nonparametrics via Neural Networks is the first book to focus on neural networks in the context of nonparametric regression and classification, working within the Bayesian paradigm. Its goal is to demystify neural networks, putting them firmly in a statistical context rather than treating them as a black box. This approach is in contrast to existing books, which tend to treat neural networks as a machine learning algorithm instead of a statistical model. Once this underlying statistical model is recognized, other standard statistical techniques can be applied to improve the model.The Bayesian approach allows better accounting for uncertainty. This book covers uncertainty in model choice and methods to deal with this issue, exploring a number of ideas from statistics and machine learning. A detailed discussion on the choice of prior and new noninformative priors is included, along with a substantial literature review. Written for statisticians using statistical terminology, Bayesian Nonparametrics via Neural Networks will lead statisticians to an increased understanding of the neural network model and its applicability to real-world problems.To illustrate the major mathematical concepts, the author uses two examples throughout the book: one on ozone pollution and the other on credit applications. The methodology demonstrated is relevant for regression and classification-type problems and is of interest because of the widespread potential applications of the methodologies described in the book.

Many products, such as foods, personal-care products, beverages, and cleaning agents, are made by mixing ingredients together. This book describes a systematic methodology for formulating such products so that they perform according to one’s goals, providing scientists and engineers with a fast track to the implementation of the methodology. Experimental Design for Formulation contains examples from a wide variety of fields and includes a discussion of how to design experiments for a mixture setting and how to fit and interpret models in a mixture setting. It also introduces process variables, the combining of mixture and nonmixture variables in a designed experiment, and the concept of collinearity and the possible problems that can result from its presence.This is a useful manual for the formulator and can also be used by a resident statistician to teach an in-house short course. Statistical proofs are largely absent, and the formulas that are presented are included to explain how the various software packages carry out the analysis. Many examples are given of output from statistical software packages, and the proper interpretation of computer output is emphasized. Other topics presented include a discussion of an effect in a mixture setting, the presentation of elementary optimization methods, and multiple-response optimization wherein one seeks to optimize more than one response.

The unlikely worlds of sports fans and statisticians collide in this interesting and accessible collection of previously-published articles on the use of statistics to analyze sports, which the editors have thoughtfully culled from a variety of American Statistical Association (ASA) publications. Heavily weighted in the areas of competition (rating players and teams, evaluating strategies for victory), the articles vary in mathematical complexity, but most will be accessible to readers with a general knowledge of statistics. Newly written material from the editors and other notable contributors introduces each section of the book, and a chapter with suggestions on using the articles in the classroom is included.Organized by sport to make it easy for readers to find the papers in their particular areas of interest, Anthology of Statistics in Sports contains separate sections devoted to the major North American team sports of baseball, football, basketball, and ice hockey. Two additional sections cover miscellaneous sports and more general issues related to sports and statistics. This book grew from the efforts of members of the ASA Section on Statistics in Sports, which is dedicated to promoting high professional standards in the application of statistics to sports and fostering statistical education in sports.

Provides a protocol for conducting gauge repeatability and reproducibility (R&R) experiments, which are required whenever a new test system is developed to monitor a manufacturing process. This protocol - not currently available in other books or technical reports - is used to determine whether the testing system is capable of monitoring the manufacturing process with the desired level of accuracy and precision. With an up-to-date summary of methods used to construct confidence intervals in normal-based random and mixed analysis of variance (ANOVA) models, this comprehensive book will be useful to scientists in all fields of application who wish to construct interval estimates for ANOVA model parameters. It includes approaches that can be applied to any ANOVA model, and because it contains detailed examples of all computations, practitioners will be able to apply the methods easily. The book describes methods for constructing two types of confidence intervals: modified large-sample (MLS) and generalized confidence intervals. Computer codes written in SAS and Excel are provided to perform the computations. Appendices are included for readers who are unfamiliar with confidence intervals or lack a basic understanding of random and mixed ANOVA models.

The analysis of means (ANOM) is a graphical procedure used to quantify differences among treatment groups in a variety of experimental design and observational study situations. The ANOM decision chart allows one to easily draw conclusions and interpret results with respect to both statistical and practical significance. It is an excellent choice for multiple comparisons of means, rates, or proportions and can be used with both balanced and unbalanced data. Key advances in ANOM procedures that have appeared only in technical journals during the last 20 years are included in this first comprehensive modern treatment of the ANOM containing all of the needed information for practitioners to understand and apply ANOM.This book contains examples from a wide variety of fields adapted from real-world applications and data with easy-to-follow, step-by-step instructions. It is front loaded, so potential ANOM users can find solutions to standard problems in the first five chapters. An appendix contains several SAS® examples showing the system’s ANOM capabilities and how SAS was used to produce selected ANOM decision charts in the book.Given these features, the lack of any other book on ANOM, and the recent inclusion of ANOM in SAS, this book will be a welcome addition to practitioners’ and statisticians’ bookshelves, where it will serve both as a primer and reference.

The Structural Representation of Proximity Matrices with MATLAB presents and demonstrates the use of functions (by way of M-files) within a MATLAB computational environment to effect a variety of structural representations for the proximity information that is assumed to be available on a set of objects. The representations included in the book have been developed primarily in the behavioral sciences and applied statistical literature (e.g., in psychometrics and classification), although interest in these topics now extends more widely to such fields as bioinformatics and chemometrics.Throughout the book, two kinds of proximity information are analyzed: one-mode and two-mode. One-mode proximity data are defined between the objects from a single set and are usually given in the form of a square symmetric matrix; two-mode proximity data are defined between the objects from two distinct sets and are given in the form of a rectangular matrix. In addition, there is typically the flexibility to allow the additive fitting of multiple structures to either the given one- or two-mode proximity information.This book is divided into three main sections, each based on the general class of representations being discussed. Part I develops linear and circular unidimensional and multidimensional scaling using the city-block metric as the major representational device. Part II discusses characterizations based on various graph-theoretic tree structures, specifically those referred to as ultrametrics and additive trees. Part III uses representations defined solely by order properties, particularly emphasizing what are called (strongly) anti-Robinson forms.

Personalized medicine is a medical paradigm that emphasizes systematic use of individual patient information to optimize that patient's health care, particularly in managing chronic conditions and treating cancer. In the statistical literature, sequential decision making is known as an adaptive treatment strategy (ATS) or a dynamic treatment regime (DTR). The field of DTRs emerges at the interface of statistics, machine learning, and biomedical science to provide a data-driven framework for precision medicine.The authors provide a learning-by-seeing approach to the development of ATSs, aimed at a broad audience of health researchers. All estimation procedures used are described in sufficient heuristic and technical detail so that less quantitative readers can understand the broad principles underlying the approaches. At the same time, more quantitative readers can implement these practices. This book:Provides the most up-to-date summary of the current state of the statistical research in personalized medicine.Contains chapters by leaders in the area from both the statistics and computer sciences fields.Contains a range of practical advice, introductory and expository materials, and case studies.