Pub Date : 2022-01-24DOI: 10.1080/00224065.2022.2035285
Yueyao Wang, Li Xu, Yili Hong, Rong Pan, Tyler H. Chang, T. Lux, Jon Bernard, L. Watson, K. Cameron
Abstract Performance variability management is an active research area in high-performance computing (HPC). In this article, we focus on input/output (I/O) variability, which is a complicated function that is affected by many system factors. To study the performance variability, computer scientists often use grid-based designs (GBDs) which are equivalent to full factorial designs to collect I/O variability data, and use mathematical approximation methods to build a prediction model. Mathematical approximation models, as deterministic methods, could be biased particularly if extrapolations are needed. In statistics literature, space-filling designs (SFDs) and surrogate models such as Gaussian process (GP) are popular for data collection and building predictive models. The applicability of SFDs and surrogates in the HPC variability management setting, however, needs investigation. In this case study, we investigate their applicability in the HPC setting in terms of design efficiency, prediction accuracy, and scalability. We first customize the existing SFDs so that they can be applied in the HPC setting. We conduct a comprehensive investigation of design strategies and the prediction ability of approximation methods. We use both synthetic data simulated from three test functions and the real data from the HPC setting. We then compare different methods in terms of design efficiency, prediction accuracy, and scalability. In our synthetic and real data analysis, GP with SFDs outperforms in most scenarios. With respect to the choice of approximation models, GP is recommended if the data are collected by SFDs. If data are collected using GBDs, both GP and Delaunay can be considered. With the best choice of approximation method, the performance of SFDs and GBD depends on the property of the underlying surface. For the cases in which SFDs perform better, the number of design points needed for SFDs is about half of or less than that of the GBD to achieve the same prediction accuracy. Although we observe that the GBD can also outperform SFDs for smooth underlying surface, GBD is not scalable to high dimensional experimental regions. Therefore, SFDs that can be tailored to high dimension and non-smooth surface are recommended especially when large numbers of input factors need to be considered in the model. This article has online supplementary materials.
{"title":"Design strategies and approximation methods for high-performance computing variability management","authors":"Yueyao Wang, Li Xu, Yili Hong, Rong Pan, Tyler H. Chang, T. Lux, Jon Bernard, L. Watson, K. Cameron","doi":"10.1080/00224065.2022.2035285","DOIUrl":"https://doi.org/10.1080/00224065.2022.2035285","url":null,"abstract":"Abstract Performance variability management is an active research area in high-performance computing (HPC). In this article, we focus on input/output (I/O) variability, which is a complicated function that is affected by many system factors. To study the performance variability, computer scientists often use grid-based designs (GBDs) which are equivalent to full factorial designs to collect I/O variability data, and use mathematical approximation methods to build a prediction model. Mathematical approximation models, as deterministic methods, could be biased particularly if extrapolations are needed. In statistics literature, space-filling designs (SFDs) and surrogate models such as Gaussian process (GP) are popular for data collection and building predictive models. The applicability of SFDs and surrogates in the HPC variability management setting, however, needs investigation. In this case study, we investigate their applicability in the HPC setting in terms of design efficiency, prediction accuracy, and scalability. We first customize the existing SFDs so that they can be applied in the HPC setting. We conduct a comprehensive investigation of design strategies and the prediction ability of approximation methods. We use both synthetic data simulated from three test functions and the real data from the HPC setting. We then compare different methods in terms of design efficiency, prediction accuracy, and scalability. In our synthetic and real data analysis, GP with SFDs outperforms in most scenarios. With respect to the choice of approximation models, GP is recommended if the data are collected by SFDs. If data are collected using GBDs, both GP and Delaunay can be considered. With the best choice of approximation method, the performance of SFDs and GBD depends on the property of the underlying surface. For the cases in which SFDs perform better, the number of design points needed for SFDs is about half of or less than that of the GBD to achieve the same prediction accuracy. Although we observe that the GBD can also outperform SFDs for smooth underlying surface, GBD is not scalable to high dimensional experimental regions. Therefore, SFDs that can be tailored to high dimension and non-smooth surface are recommended especially when large numbers of input factors need to be considered in the model. This article has online supplementary materials.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2022-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82735014","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-01-10DOI: 10.1080/00224065.2023.2180457
Shuxiang Fan, S. Wong, J. Zidek
Abstract When assessing the strength of sawn lumber for use in engineering applications, the sizes and locations of knots are an important consideration. Knots are the most common visual characteristics of lumber, that result from the growth of tree branches. Large individual knots, as well as clusters of distinct knots, are known to have strength-reducing effects. However, industry grading rules that govern knots are informed by subjective judgment to some extent, particularly the spatial interaction of knots and their relationship with lumber strength. This case study reports the results of an experiment that investigated and modeled the strength-reducing effects of knots on a sample of Douglas Fir lumber. Experimental data were obtained by taking scans of lumber surfaces and applying tensile strength testing. The modeling approach presented incorporates all relevant knot information in a Bayesian framework, thereby contributing a more refined way of managing the quality of manufactured lumber.
{"title":"Knots and their effect on the tensile strength of lumber: A case study","authors":"Shuxiang Fan, S. Wong, J. Zidek","doi":"10.1080/00224065.2023.2180457","DOIUrl":"https://doi.org/10.1080/00224065.2023.2180457","url":null,"abstract":"Abstract When assessing the strength of sawn lumber for use in engineering applications, the sizes and locations of knots are an important consideration. Knots are the most common visual characteristics of lumber, that result from the growth of tree branches. Large individual knots, as well as clusters of distinct knots, are known to have strength-reducing effects. However, industry grading rules that govern knots are informed by subjective judgment to some extent, particularly the spatial interaction of knots and their relationship with lumber strength. This case study reports the results of an experiment that investigated and modeled the strength-reducing effects of knots on a sample of Douglas Fir lumber. Experimental data were obtained by taking scans of lumber surfaces and applying tensile strength testing. The modeling approach presented incorporates all relevant knot information in a Bayesian framework, thereby contributing a more refined way of managing the quality of manufactured lumber.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2022-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85147875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-30DOI: 10.1080/00224065.2020.1764418
Shuai Huang
As a sister book of the book Probability by the same author, this book is supposed to be the second course in a mathematical statistics sequence of classes. The readers should have learned calculus and completed a calculusbased course in probability. As with most mathematical statistics textbooks, point estimations, interval estimation, and hypothesis testing are the core concepts. This book is particularly written for students who would have their first exposure to mathematical statistics, so the author carefully selected his materials and had focused on the understanding of statistics such as the sample mean and sample variance being also random variables as well. R is used throughout the text for graphics, computation, and Monte Carlo simulation. The homework is comprehensive. From all these aspects, this book has a similar style as the other book of Probability by the same author. The book’s organization is deceptively simple: it only has four chapters. Chapter 1, almost 100 pages, is about random sampling. Chapter 2, another 100 pages, is about point estimation. Chapter 3, 135 pages, is about interval estimation. Chapter 4, 133 pages, is about hypothesis testing. This “simple” structure makes the four pillars of mathematical statistics very clear to readers who first learn the topic. Within each chapter, just like in the book Probability, each concept is presented in detail and in multiple aspects. And when calculation is involved, enough middle steps are preserved so readers can easily follow the steps. One notable example is the presentation of the hypothesis testing. Not like many other textbooks that start with proven methods such as the Z-test, this book introduces the big picture first, and this big picture includes “a hunch”: it presents in the very beginning a clear outline of the 12 steps for hypothesis testing, starts with “a hunch, or theory, concerning a problem of interest,” then moves to the second step “translate the theory into a question concerning an unknown parameter theta,” then “state the null hypothesis of theta.” ... Then technical explanation of many of these steps is given in detail. The type 1 and type 2 errors are also presented right along with this 12-step outline. What is more, strange (i.e., idiosyncratic) forms of hypothesis testing are presented! It concerns three brothers, Chico, Harpo, and Groucho. Each of them comes up with their own testing statistics, e.g., x1þ x2, min (x1, x2), or max(x1, x2), where x1 and x2 are random samples of size 2 from a uniform distribution U(0, theta). Is theta 1⁄4 5, or theta 1⁄4 2? Is this an allusion to the three little pigs? Nonetheless, this is a hilarious example that very effectively instructs the technical details of hypothesis testing, but also revives this “ancient” technique that tells readers that, in using the proven hypothesis testing methods, we actually have made choices (i.e., each of the three brothers’ proposals have pros and cons, in terms of the type 1 and
{"title":"Mathematical Statistics","authors":"Shuai Huang","doi":"10.1080/00224065.2020.1764418","DOIUrl":"https://doi.org/10.1080/00224065.2020.1764418","url":null,"abstract":"As a sister book of the book Probability by the same author, this book is supposed to be the second course in a mathematical statistics sequence of classes. The readers should have learned calculus and completed a calculusbased course in probability. As with most mathematical statistics textbooks, point estimations, interval estimation, and hypothesis testing are the core concepts. This book is particularly written for students who would have their first exposure to mathematical statistics, so the author carefully selected his materials and had focused on the understanding of statistics such as the sample mean and sample variance being also random variables as well. R is used throughout the text for graphics, computation, and Monte Carlo simulation. The homework is comprehensive. From all these aspects, this book has a similar style as the other book of Probability by the same author. The book’s organization is deceptively simple: it only has four chapters. Chapter 1, almost 100 pages, is about random sampling. Chapter 2, another 100 pages, is about point estimation. Chapter 3, 135 pages, is about interval estimation. Chapter 4, 133 pages, is about hypothesis testing. This “simple” structure makes the four pillars of mathematical statistics very clear to readers who first learn the topic. Within each chapter, just like in the book Probability, each concept is presented in detail and in multiple aspects. And when calculation is involved, enough middle steps are preserved so readers can easily follow the steps. One notable example is the presentation of the hypothesis testing. Not like many other textbooks that start with proven methods such as the Z-test, this book introduces the big picture first, and this big picture includes “a hunch”: it presents in the very beginning a clear outline of the 12 steps for hypothesis testing, starts with “a hunch, or theory, concerning a problem of interest,” then moves to the second step “translate the theory into a question concerning an unknown parameter theta,” then “state the null hypothesis of theta.” ... Then technical explanation of many of these steps is given in detail. The type 1 and type 2 errors are also presented right along with this 12-step outline. What is more, strange (i.e., idiosyncratic) forms of hypothesis testing are presented! It concerns three brothers, Chico, Harpo, and Groucho. Each of them comes up with their own testing statistics, e.g., x1þ x2, min (x1, x2), or max(x1, x2), where x1 and x2 are random samples of size 2 from a uniform distribution U(0, theta). Is theta 1⁄4 5, or theta 1⁄4 2? Is this an allusion to the three little pigs? Nonetheless, this is a hilarious example that very effectively instructs the technical details of hypothesis testing, but also revives this “ancient” technique that tells readers that, in using the proven hypothesis testing methods, we actually have made choices (i.e., each of the three brothers’ proposals have pros and cons, in terms of the type 1 and ","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85795417","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-30DOI: 10.1080/00224065.2021.2019569
Caleb King, T. Bzik, Peter A. Parker, M. Wells, Benjamin R. Baer
{"title":"Addendum to “Estimating pure-error from near replicates in design of experiments”","authors":"Caleb King, T. Bzik, Peter A. Parker, M. Wells, Benjamin R. Baer","doi":"10.1080/00224065.2021.2019569","DOIUrl":"https://doi.org/10.1080/00224065.2021.2019569","url":null,"abstract":"","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76614565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-12-08DOI: 10.1080/00224065.2021.2006583
Jing Li
In Industrial Data Analytics for Diagnosis and Prognosis A Random Effects Modelling Approach, distinguished engineers Shiyu Zhou and Yong Chen deliver a rigorous and practical introduction to the random effects modeling approach for industrial system diagnosis and prognosis. In the book’s two parts, general statistical concepts and useful theory are described and explained, as are industrial diagnosis and prognosis methods. The accomplished authors describe and model fixed effects, random effects, and variation in univariate and multivariate datasets and cover the application of the random effects approach to diagnosis of variation sources in industrial processes. They offer a detailed performance comparison of different diagnosis methods before moving on to the application of the random effects approach to failure prognosis in industrial processes and systems.
{"title":"Industrial Data Analytics for Diagnosis and Prognosis: A Random Effects Modeling Approach","authors":"Jing Li","doi":"10.1080/00224065.2021.2006583","DOIUrl":"https://doi.org/10.1080/00224065.2021.2006583","url":null,"abstract":"In Industrial Data Analytics for Diagnosis and Prognosis A Random Effects Modelling Approach, distinguished engineers Shiyu Zhou and Yong Chen deliver a rigorous and practical introduction to the random effects modeling approach for industrial system diagnosis and prognosis. In the book’s two parts, general statistical concepts and useful theory are described and explained, as are industrial diagnosis and prognosis methods. The accomplished authors describe and model fixed effects, random effects, and variation in univariate and multivariate datasets and cover the application of the random effects approach to diagnosis of variation sources in industrial processes. They offer a detailed performance comparison of different diagnosis methods before moving on to the application of the random effects approach to failure prognosis in industrial processes and systems.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89676795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-11-01DOI: 10.1080/00224065.2021.1991863
Qi Zhou, William Li, Hongquan Xu
Abstract Extensive studies have been conducted on how to select efficient designs with respect to a criterion. Most design criteria aim to capture the overall efficiency of the design across all columns. When prior information indicated that a small number of factors and their two-factor interactions (2fi’s) are likely to be more significant than other effects, commonly used minimum aberration designs may no longer be the best choice. Motivated by a real-life experiment, we propose a new class of regular fractional factorial designs that focus on estimating a subset of columns and their corresponding 2fi’s clear of other important effects. After introducing the concept of individual clear effects (iCE) to describe clear 2fi’s involving a specific factor, we define the clear effect pattern criterion to characterize the distribution of iCE’s over all columns. We then obtain a new class of designs that sequentially maximize the clear effect pattern. These newly constructed designs are often different from existing optimal designs. We develop a series of theoretical results that can be particularly useful for constructing designs with large run sizes, for which algorithmic construction becomes computationally challenging. We also provide some practical guidelines on how to choose appropriate designs with respect to different run size, the number of factors, and the number of 2fi’s that need to be clear.
{"title":"Utilizing individual clear effects for intelligent factor allocations and design selections","authors":"Qi Zhou, William Li, Hongquan Xu","doi":"10.1080/00224065.2021.1991863","DOIUrl":"https://doi.org/10.1080/00224065.2021.1991863","url":null,"abstract":"Abstract Extensive studies have been conducted on how to select efficient designs with respect to a criterion. Most design criteria aim to capture the overall efficiency of the design across all columns. When prior information indicated that a small number of factors and their two-factor interactions (2fi’s) are likely to be more significant than other effects, commonly used minimum aberration designs may no longer be the best choice. Motivated by a real-life experiment, we propose a new class of regular fractional factorial designs that focus on estimating a subset of columns and their corresponding 2fi’s clear of other important effects. After introducing the concept of individual clear effects (iCE) to describe clear 2fi’s involving a specific factor, we define the clear effect pattern criterion to characterize the distribution of iCE’s over all columns. We then obtain a new class of designs that sequentially maximize the clear effect pattern. These newly constructed designs are often different from existing optimal designs. We develop a series of theoretical results that can be particularly useful for constructing designs with large run sizes, for which algorithmic construction becomes computationally challenging. We also provide some practical guidelines on how to choose appropriate designs with respect to different run size, the number of factors, and the number of 2fi’s that need to be clear.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78446053","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-28DOI: 10.1080/00224065.2021.1991251
Zhijun Wang, Chunjie Wu, Miaomiao Yu, F. Tsung
Abstract Conventional self-starting control schemes can perform poorly when monitoring processes with early shifts, being limited by the number of historical observations sampled. In real applications, pre-observed data sets from other production lines are always available, prompting us to propose a scheme that monitors the target process using historical data obtained from other sources. The methodology of self-taught clustering from unsupervised transfer learning is revised to transfer knowledge from previous observations and improve out-of-control (OC) performance, especially for processes with early shifts. However, if the difference in distribution between the target process and the pre-observed data set is large, our scheme may not be the best. Simulation results and two illustrative examples demonstrate the superiority of the proposed scheme.
{"title":"Self-starting process monitoring based on transfer learning","authors":"Zhijun Wang, Chunjie Wu, Miaomiao Yu, F. Tsung","doi":"10.1080/00224065.2021.1991251","DOIUrl":"https://doi.org/10.1080/00224065.2021.1991251","url":null,"abstract":"Abstract Conventional self-starting control schemes can perform poorly when monitoring processes with early shifts, being limited by the number of historical observations sampled. In real applications, pre-observed data sets from other production lines are always available, prompting us to propose a scheme that monitors the target process using historical data obtained from other sources. The methodology of self-taught clustering from unsupervised transfer learning is revised to transfer knowledge from previous observations and improve out-of-control (OC) performance, especially for processes with early shifts. However, if the difference in distribution between the target process and the pre-observed data set is large, our scheme may not be the best. Simulation results and two illustrative examples demonstrate the superiority of the proposed scheme.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82710881","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-26DOI: 10.1080/00224065.2023.2196034
M. Ebadi, Shoja'eddin Chenouri, Stefan H. Steiner
Abstract One of the significant challenges in monitoring the quality of products today is the high dimensionality of quality characteristics. In this paper, we address Phase I analysis of high-dimensional processes with individual observations when the available number of samples collected over time is limited. Using a new charting statistic, we propose a robust procedure for parameter estimation in Phase I. This robust procedure is efficient in parameter estimation in the presence of outliers or contamination in the data. A consistent estimator is proposed for parameter estimation and a finite sample correction coefficient is derived and evaluated through simulation. We assess the statistical performance of the proposed method in Phase I. This assessment is carried out in the absence and presence of outliers. We show that, in both cases, the proposed control chart scheme effectively detects various kinds of shifts in the process mean. Besides, we present two real-world examples to illustrate the applicability of our proposed method.
{"title":"Phase I analysis of high-dimensional processes in the presence of outliers","authors":"M. Ebadi, Shoja'eddin Chenouri, Stefan H. Steiner","doi":"10.1080/00224065.2023.2196034","DOIUrl":"https://doi.org/10.1080/00224065.2023.2196034","url":null,"abstract":"Abstract One of the significant challenges in monitoring the quality of products today is the high dimensionality of quality characteristics. In this paper, we address Phase I analysis of high-dimensional processes with individual observations when the available number of samples collected over time is limited. Using a new charting statistic, we propose a robust procedure for parameter estimation in Phase I. This robust procedure is efficient in parameter estimation in the presence of outliers or contamination in the data. A consistent estimator is proposed for parameter estimation and a finite sample correction coefficient is derived and evaluated through simulation. We assess the statistical performance of the proposed method in Phase I. This assessment is carried out in the absence and presence of outliers. We show that, in both cases, the proposed control chart scheme effectively detects various kinds of shifts in the process mean. Besides, we present two real-world examples to illustrate the applicability of our proposed method.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80554616","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-21DOI: 10.1080/00224065.2021.1991250
Richard N. McGrath, Baffour Koduah
Abstract A popular approach for estimating location and dispersion effects in replicated designs under the common assumption of normal and independent errors is to use two linked generalized linear models (glms). This approach uses an asymptotic estimate for the variance of dispersion effect estimates and is very sensitive to the normality assumption. It is also possible to identify dispersion effects (after a logarithmic transformation) by using methods developed for identifying location effects in unreplicated designs. One such method is rather robust to the normality assumption but lacks power relative to the glm approach. We introduce a hybrid approach that strikes a balance between power and robustness when used for dispersion effect identification.
{"title":"Powerful and robust dispersion contrasts for replicated orthogonal designs","authors":"Richard N. McGrath, Baffour Koduah","doi":"10.1080/00224065.2021.1991250","DOIUrl":"https://doi.org/10.1080/00224065.2021.1991250","url":null,"abstract":"Abstract A popular approach for estimating location and dispersion effects in replicated designs under the common assumption of normal and independent errors is to use two linked generalized linear models (glms). This approach uses an asymptotic estimate for the variance of dispersion effect estimates and is very sensitive to the normality assumption. It is also possible to identify dispersion effects (after a logarithmic transformation) by using methods developed for identifying location effects in unreplicated designs. One such method is rather robust to the normality assumption but lacks power relative to the glm approach. We introduce a hybrid approach that strikes a balance between power and robustness when used for dispersion effect identification.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83323856","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-20DOI: 10.1080/00224065.2022.2034487
S. Knoth, Nesma A. Saleh, Mahmoud A. Mahmoud, W. Woodall, V. Tercero-Gómez
Abstract Many extensions and modifications have been made to standard process monitoring methods such as the exponentially weighted moving average (EWMA) chart and the cumulative sum (CUSUM) chart. In addition, new schemes have been proposed based on alternative weighting of past data, usually to put greater emphasis on past data and less weight on current and recent data. In other cases, the output of one process monitoring method, such as the EWMA statistic, is used as the input to another method, such as the CUSUM chart. Often the recursive formula for a control chart statistic is itself used recursively to form a new control chart statistic. We find the use of these ad hoc methods to be unjustified. Statistical performance comparisons justifying the use of these methods have been either flawed by focusing only on zero-state run length metrics or by making comparisons to an unnecessarily weak competitor.
{"title":"A critique of a variety of “memory-based” process monitoring methods","authors":"S. Knoth, Nesma A. Saleh, Mahmoud A. Mahmoud, W. Woodall, V. Tercero-Gómez","doi":"10.1080/00224065.2022.2034487","DOIUrl":"https://doi.org/10.1080/00224065.2022.2034487","url":null,"abstract":"Abstract Many extensions and modifications have been made to standard process monitoring methods such as the exponentially weighted moving average (EWMA) chart and the cumulative sum (CUSUM) chart. In addition, new schemes have been proposed based on alternative weighting of past data, usually to put greater emphasis on past data and less weight on current and recent data. In other cases, the output of one process monitoring method, such as the EWMA statistic, is used as the input to another method, such as the CUSUM chart. Often the recursive formula for a control chart statistic is itself used recursively to form a new control chart statistic. We find the use of these ad hoc methods to be unjustified. Statistical performance comparisons justifying the use of these methods have been either flawed by focusing only on zero-state run length metrics or by making comparisons to an unnecessarily weak competitor.","PeriodicalId":54769,"journal":{"name":"Journal of Quality Technology","volume":null,"pages":null},"PeriodicalIF":2.5,"publicationDate":"2021-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76906948","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：