Bayesian Data Analysis

By Andrew Gelman, J. Carlin, Hal S. Stern, David B. Dunson, Aki Vehtari, Donald B. Rubin. Technometrics, 2004

5,883 citations recorded by the source corpus.

trol charts, and their relationship to process capability. Chapter 7, “Preparing for Control Charts,” presents procedures for determining sample size and frequency, with some comparison of variable and attribute control charts. All of this is generally informative, but we are still waiting to find out how to produce a control chart. Chapter 8, “Variables Charts,” explains how to do this. It presents 13 steps for producing Xbar and R charts, including 15 figures illustrating various process control situations. However, step 9, “analyze the range chart for process control,” refers to “runs” and “trends” that have not yet been defined, as well as the “1/3 or 2/3 rule,” which is not described until much later. This chapter also presents X and moving-range charts, and Xbar and S charts briefly. Cusum charts are mentioned but dismissed as “not very convenient to use on a routine basis.” It would have been better to exclude them altogether, because not enough information is presented to allow the reader to make a cusum chart. Chapter 9, “Control Charts for Attributes,” presents detailed steps for producing p charts, np charts, c charts, and u charts, as well as a brief discussion on D charts, for weighted number of defects. Chapter 10, “Other Charts,” presents steps for producing median and range charts, with the same 15 figures presented in Chapter 8 for Xbar and R charts. Then it discusses standardized charts for attribute data, including a “yield (q) chart.” This term was not used previously in the book. “Difference from reference” control charts are discussed briefly, followed by “the lot plot method,” for which no motivation is provided. Then a “modified control chart” (which sets the control limits out from an interval on the mean) is described, but the presentation is so brief and cryptic as to be useless. The chart for tool wear, with the center adjusted over time, is fairly well described, although the example calculations for the regression equation are incomplete. There is a good discussion of precontrol. Chapter 11, “Control Chart Signals,” is useful, but would have been better placed earlier in the book. Chapter 12, “Process Control and Capability,” contrasts these two concepts and presents a procedure for evaluating process capability and estimating capability indices. This is a generally good presentation but with occasional missteps, such as “another advantage of knowing the Cp and Cpk is that once these values are known, one can determine the shape of the distribution” (p. 258), and a reference to the “tolerance zone” (p. 263), which is not defined. I found Chapter 13, “Short-Run SPC,” a confusing “cookbook” presentation of how to produce these charts, without adequate background or justification for the process. It presents 11 different types of variables charts with no discussion of when each chart might be appropriate. This chapter really needs better organization and discussion. Chapter 14, “Distribution Shape and Stability,” suggests that asymmetric control limits might be appropriate in some cases and presents calculation of control limits for the exponential distribution. This is generally useful and well presented, as is a discussion of trending processes. Chapter 15, “The Measurement Process,” discusses repeatability, reproducibility, stability, accuracy, precision, and bias, with steps for performing an R&R study. Again, the material is generally useful and well presented, although the analysis of variance results has insufficient accompanying explanation and should have been excluded. Chapter 16, “Machine Acceptance Overview,” is supposed to “bring everything together” (p. 363), but it seems pretty useless. However, it does include instructions for installing Minitab on a PC from floppy discs. Finally, Chapter 17, “SPC in Nonmanufacturing,” compares criteria for success and quality characteristics in manufacturing and service industries and identifies tools that might be used in the service industry, followed by a discussion of an inventory control program. I guess it was left over from another book. Although much of the material on producing control charts is generally well presented for the Six Sigma practitioner, the book suffers from the inclusion of information on a range of topics that do not belong. In addition, the presentation of information on some important topics (short-run SPC in particular) is too brief. Overall, this book does not stand on its own as a reference for statistical process control.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/936e57003ad0c6ca4aac4eda4068d2254dbf2ecf

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Bayesian inference for long-term prediction of significant wave height (2007, background): “Examples of its most direct use within extreme value analysis include: Smith and Naylor (1987) who compare Bayesian and maximum likelihood estimators for the Weibull distribution; Lingappaiah (1984), Achcar et al.”
Measuring long‐distance seed dispersal in complex natural environments: an evaluation and integration of classical and genetic methods (2008, methodology): “FITT ING THE MODELS TO EMPIRICAL DATA We used hierarchical Bayesian models to fit the models to the empirical data set (Gelman et al. 1995; Clark 2007).”
Combining Social Media and Survey Data to Nowcast Migrant Stocks in the United States (2020, methodology): “Standard diagnostic checks using trace plots and the Gelman and Rubin diagnostic were used to check convergence (Gelman et al., 2013).”
Fine-scale distribution of juvenile Atlantic and Mediterranean loggerhead turtles (Caretta caretta) in the Mediterranean Sea (2014, methodology): “Iterated chains were only considered reliable when the gelman–rubin criterion was fulfilled (g-r shrink factor <1.2 for all parameters; gelman et al. 1996).”