Statistics for Experimenters - Second Edition

Topic: Statistics, Six Sigma, Management Improvement

The classic Statistics for Experimenters has been updated by George Box and Stu Hunter, two of the three original authors. Bill Hunter, who was my father, and the other author, died in 1986. Order online: Statistics for Experimenters: Design, Innovation, and Discovery , 2nd Edition by George E. P. Box, J. Stuart Hunter, William G. Hunter. I happen to agree with those who call this book a classic, however, I am obviously biased.

Google Scholar citations for the first edition of Statistics for Experimenters.
Citations in Cite Seer to the first edition.

The first edition includes the text of Experiment by Cole Porter. In 1978 finding a recording of this song was next to impossible. Now Experiment can be heard on the De-Lovely soundtrack.

Text from the publisher on the 2nd Edition:
Rewritten and updated, this new edition of Statistics for Experimenters adopts the same approaches as the landmark First Edition by teaching with examples, readily understood graphics, and the appropriate use of computers. Catalyzing innovation, problem solving, and discovery, the Second Edition provides experimenters with the scientific and statistical tools needed to maximize the knowledge gained from research data, illustrating how these tools may best be utilized during all stages of the investigative process. The authors’ practical approach starts with a problem that needs to be solved and then examines the appropriate statistical methods of design and analysis.

Providing even greater accessibility for its users, the Second Edition is thoroughly revised and updated to reflect the changes in techniques and technologies since the publication of the classic First Edition.

Among the new topics included are:

• Graphical Analysis of Variance
• Computer Analysis of Complex Designs
• Simplification by transformation
• Hands-on experimentation using Response Service Methods
• Further development of robust product and process design using split plot arrangements and minimization of error transmission
• Introduction to Process Control, Forecasting and Time Series
• Illustrations demonstrating how multi-response problems can be solved using the concepts of active and inert factor spaces and canonical spaces
• Bayesian approaches to model selection and sequential experimentation

An appendix featuring Quaquaversal quotes from a variety of sources including noted statisticians and scientists to famous philosophers is provided to illustrate key concepts and enliven the learning process.

All the computations in the Second Edition can be done utilizing the statistical language R. Functions for displaying ANOVA and lamba plots, Bayesian screening, and model building are all included and R packages are available online. All theses topics can also be applied utilizing easy-to-use commercial software packages.

Complete with applications covering the physical, engineering, biological, and social sciences, Statistics for Experimenters is designed for individuals who must use statistical approaches to conduct an experiment, but do not necessarily have formal training in statistics. Experimenters need only a basic understanding of mathematics to master all the statistical methods presented. This text is an essential reference for all researchers and is a highly recommended course book for undergraduate and graduate students.

post by John Hunter

Curious Cat Articles and Links

We have created a new blog, Curious Cat Articles and Links, where we will post links to articles and web sites we find interesting.

Control Charts in Health Care

Topic: Management Improvement

This post is an edtied version of a message I sent to the Deming Electronic Network.

I find the “control charts in health care” thread quite interesting.

From Mike Woolbert's post
> I have read many comments about the 8 minute ambulance trip.
> This doesn't seem to be a system measure, but a result measure.

It seems to me the 8 minute (90% of the time) measure is an attempt at a process measure (in a sense, you can see it as a result measure, but it is also a measure that will have an impact on overall results and as such can be used a process indicator). For it to be a process measure rather than than a process target however, it should actual be a measure of what has happened not a statement that we want to have 90% arrive within 8 minutes.

Jonathan Siegel's comments on this topic were excellent.

The control chart was developed to aid in process improvement. A control chart helps monitor the process (to aid in putting in place counter-measures, when needed, and for identification of special causes). The control chart can be used to see if the process is in control and what the expected results from the system are.

The 90% in 8 minutes measure seems to me to be the desire or goal (as a proxy for a “customer specification” or “voice of the customer”). The control chart will show what the process is capable of: “voice of the process.” There could be another voice of the process that excludes the 10% longest times, but I don't think that would be a control chart, it would be something else. I think that is fine, as long as what it was, and is not, is understood. Maybe there is some way to have the 90% measures as a control chart that I just don't think of it right now.

Learning that the process is in control (using a control chart) can help guide the improvement efforts. When the process is in control certain process improvement strategies are likely to be effective. And PDSA pilots can be analyzed based on the knowledge that the system is in control. I think Brian Joiner's 4th Generation Management has very good information on this topic, see pages 140-153. Essentially, focus on improving the process as a whole, using PDSA and experimentation, rather than focusing on special causes. In addition, those pages talk about stratifying and disagragating the data to help focus improvement efforts.

If the system is not in control different improvement strategies are most likely to be effective. Again 4th Generation Management has good ideas pages 138-140. Essentially, in this case the type of management most often seen, find causes and react is often the best strategy. The long term strategy should be to bring the process into control and then improve the process.

Thinking about the ambulance response process I have the following thoughts.

It seems likely the 8 minute figure is based on experience such that the system outcomes (the health of the patient) will decline as times increase, which makes sense to me. And possibly the 8 minute figure is based on some studies that make that a sensible figure to use. But I would want to get the data (both looking at the literature overall and viewing internal records), on why 8 minutes versus 6 or 10.

Knowing how important the 8 minute target is may be critical to improvement strategies. The point of using a control chart, and many of the management improvement tools, are to improve the efficiency and effectiveness of resources spent improving. The trick is not really to improve (that is pretty easy) the trick is to improve quickly and effectively (and in a competitive marketplace to improve more quickly than competitors). Where improvement resources are targeted is critical. In deciding which improvement options to explore it is important to understand the impact on the outcome (in this case the health of the patient).

As stated in an earlier message, the real aim is improving the outcome for the patient. Assuming that outcomes decline as time increases. One way to do this is by finding and eliminating special causes (or adding them in case the special causes are beneficial).

I think the control chart would need to include all data not just 90%. The control chart is a measure of the performance of the process. The 90% within 8 minutes seems more

This message could benefit from more thought and editing on my part but I have to run so I will send it as it is and hope it is useful - even in this rough form.

John Hunter
www.johnhunter.com

Baby Names - Web Application

Topic: Internet

Cool site: Baby name wizard - name popularity java application

Great example of a traffic generating tool/content use to help market a product. The aplication allows visitors to view the popularity of names over time. The site provides a useful and interesting service for free and in turn get the oportunity to sell their product (a book on baby names - The Baby Name Wizard).

This is a great example of what the internet can do well - providing something of value to visitors and serving as a marketing vehicle for the organization. And by providing something interesting the site gets free publicity (such as this post).

Private Consumer Information Stolen from ChoicePoint

Topics: Economics, Privacy, Government

According to an article on CNN.com, "tens of thousands of U.S. consumers face a greater risk of identity theft after criminals gained access to a database of personal records compiled by ChoicePoint Inc."

"In California, the only state that requires companies to disclose security breaches, ChoicePoint sent warning letters to 30,000 to 35,000 consumers advising them to check their credit reports."

"Chris Hoofnagle, associate director with the Electronic Privacy Information Center... This is a prime example of how they don't and why ChoicePoint should be subject to federal privacy regulations," He is completely right.

More information is detailed in the letter sent by EPIC to the Federal Trade Commission on 1 Feb 2005: Request for investigation into data broker products for compliance with the Fair Credit Reporting Act.

It is a shame the government is failing to serve their role, as regulator, as necessary in a properly functioning capitalist economy. At least California has taken the minimal step of requiring the companies to notify consumers when the companies discover they have made their private information available improperly. The failure to regulate in areas with negative economic externalities causes great damage to the economy and the individuals of the society. Unfortunately the government seems uninterested in these responsibilities. For that reason citizens are subject to negative externalities that they should not have to suffer. If government leaders properly played their role as regulator.

I don't like the idea that my private information is held by companies that continue to fail to properly protect that private information. Their right to maintain records of private information needs to be much more effectively regulated. And citizen's should be given much more control over the flow of their private information. I wonder how much money the politicians get from those who trade in private information? I imagine, the money recieved might provide a compelling explanation for the failure to regulate properly to those who are sceptical of the pure intentions of our "representatives."

Updated 17 Feb
Dan Gilmor has added a post on this topic - A Dossier on Your Life: Now Criminals May Have It. My comments on that post:

Unfortunately many in the United States of American have come to equate no regulations with capitalism. Government has a critical role to play in regulating the market in capitalist theory. However, many political leaders don't understand the basic tenets of capitalism and fail to fulfil their vital role properly.

Europe is way ahead on protecting privacy rights for citizens

Management Improvemnet Articles

Topic: Management - Library Additions

Recent additions to the Curious Cat Management Improvement Library include:

• Six Sigma for the Little Guy by Mikel J. Harry and J. Douglas Crawford
• What are the Measures That Matter by Art Kleiner
• Critical Chain Scheduling and Buffer Management by Frank Patrick
• Theory of Constraints by Sid Sytsma
  
• Tom Peters Hates Quality but Loves Its Principles and Practices by John Pryor
  

Find links to these, and other new additions, on the Curious Cat Management Improvement New Articles Page or search for management improvement articles.

Web Search Improvements

Topic: Internet - Search Improvement

I have been waiting for Google to provide a way around an obvious problem with their search results for a long time. They choose to provide results that do not display matching content when you click on the result link. Since this was such an obvious failure to meet the needs of web search customers I figured a solution would be quick.

After a time I decided I should tell them I don't want "results" that don't display the content when I click on the link results Google provides. So I suggested, I, and many others, would like a way of indicating we don't want those results returned to us. It has now been more than 6 months and still nothing. I get an increasing number of these bad results. Combined with the improvement great improvement of Yahoo and MSN searches I am now willing to give them a try as my primary search and see how they preform.

In this case, I am not referring to pages where the site owner has deceived Google by providing false content to Google's bot. For those cases, see improvement suggestion 2, below. What I am talking about are pages, such as:

Google search result link goes to this page which has no relevant content.

Link to the page shown in the image. Where Google, I have to believe, knows has the content not available from the search result they provide but hidden away behind free, or paid, registration. I can understand some people might want such items returned. I can't imagine people would choose that as the default option but even if Google wanted that to be the default option that would be fine with me as long as they provided a way for me to only get search results that returned links that linked to a page with the content Google found to match my search. It appears Yahoo! doesn't provide this option either. If there is a way to accomplish what I want that you know of please let me know.

Improvements I would like to see:

1. Ability to restrict results so only those links that will display the content are returned.
2. Ability to restrict sites from the search results in my preferences (Google doesn't seem to be able to restrict certain obvious spam names databases from results, so let me do so myself. It would be best if there could be filter lists that are maintained by various communities that I could choose to apply. I can imagine there are technical hurdles to overcome to make this a reality).
3. Ability to restrict types of results by file type in my preference (such as power point or pdf files). I would like to be able to have more obvious indications when the file is one of those types. Again I can appreciate this might not be what everyone wants. I have no problem with the current default option remaining, I would just like the ability to set my preferences and have those used for my search results.
4. Easier way to have results restricted to certain sites (including site lists). It would be nice the search engine could retain a list of sites I have used in the past and let me select from those. Also I would like to be able to select multiple sites. And I would like to have groups of sites (site lists - so I could just select, for example, the "management improvement" sites) and the search results would be filtered only for those have included in my list of management sites.
5. With the ability to filter results list above, an appropriate interface to make such options easy to use on the search result page. The current preference page works ok, for its purpose of setting long term preferences. It is not very good for one time search changes and certainly would not work well for quickly manipulating the results based on the options shown above. I would imagine smart, talented people could design a toolbar that would meet this need well.
   

Theory of Knowledge

Deming's Management System, as expressed in his book: The New Economics has four interdependent parts:

• Appreciation for a System (systems thinking)
  
• Knowledge about Variation (see: variation definition)
• Theory of Knowledge
• Psychology (the human element of management systems)

Quotes on the Theory of Knowledge portion from The New Economics by W. Edwards Deming.

• "Management is Prediction. The theory of knowledge that management in any form is prediction" (Page 101)
  
• "Knowledge is built upon theory... Rational prediction requires theory and build knowledge through systematic revision and comparison of theory based on comparison of prediction with observation." (Page 102).
  

It took me many years to appreciate the improtance of prediction and theory to aid in learning and improvement. When managing many fail to predict when attempting to test improvement ideas through what should be experiments (often they are just changes without verification the change produced a desired effect, any learning or study of the results of the change). Without prediction learning is much less (if there is any at all) than it would be with such prediction.

In addition, when it is understood that management is based on prediction then the impact of the other 3 areas of the system of profound knowledge are clearer.

By exploring the basis for the prediction one must understand the theory they are operating with to make the prediction. Most often managers fail to develop a theory that allows them to predict. They fail to predict the results of an attempt to improve (PDSA), they fail to analyze the results of that improvement experiment, they fail to learn about the system that they are managing (since they fail to predict and then learn) and therefore cannot refine their theory based on their learning. When failing to do these things it is a surprize that learning is very ineffective? And without learning how can effective improvement be expected?

With, even a fairly simple understanding of the theory of knowledge the effectiveness of management improvement efforts are greatly increased. This topic is difficult for most to understand, I recommend reading chapter four of the New Economics. And I recommend returning to that chapter periodically as you apply management improvement techniques and learn and grow as a manager.

More from the Curious Cat Management Glossary: Theory of Knowledge