iSixSigma

Are Six Sigma stats old fashioned and flawed?

Six Sigma – iSixSigma Forums Old Forums General Are Six Sigma stats old fashioned and flawed?

Viewing 75 posts - 1 through 75 (of 75 total)
  • Author
    Posts
  • #39172

    Six Sigma Tom
    Member

    It seems to me that the statistical tools and techniques traditionally used in Six Sigma are way out of date.- They were designed in an era when data were scarce and computing power was expensive. We now have gigantic data warehouses and everybody has a super computer on their desk.- They were designed for mass production manufacturing processes. We now have transactional processes, high variety production, short runs, etc.- The traditional way of teaching statistics- statistical hypothesis inference testing- (supply your own acronym) has been discredited by leading statisticians for a variety of reasons.This has all been known for a long time. There are better ways. What’s holding Six Sigma back?Six Sigma Tom

    0
    #118661

    New VGB
    Participant

    Agree fully with you.Efforts should  be  carried  out  to modify SS stats as you have  mentioned,regards

    0
    #118672

    Royale with Cheese
    Member

    Fine, you think the statitistical tools are out of date. What does that have to do with effective change management and leadership development?
    The statistical tools I learned in training were just a starting point…If someone taught me some new fan-tangled way to analyze data, well I’d just throw that in the toolbox with the other stuff I learned and keep right on going. Whats your point?

    0
    #118673

    Robert Butler
    Participant

    The statistical tools taught by Six Sigma are not old fashioned, way out of date, or flawed.  They are, to be sure, elementary but they meet the requirement for brevity (with respect to effort needed for their presentation) and utility (with respect to their general level of applicability).  Both of these are essential from the standpoint of initial training.  A 4 week training course means the initial training in statistical methods is two weeks in duration. At the end of training a typical trainee will have something less than two weeks worth of understanding of statistical methods and that typical trainee will be expected to go out in the workplace and actually do something with the things they have learned.
      The complexity of statistics means if you are going to have any hope of giving someone tools in the time allotted these tools will have to be simple and they will have to be taught with a lot of boilerplate about do’s and don’ts. It must never be forgotten (although it appears this is often not the case) that while this boilerplate is acceptable it is an oversimplification and it must always be questioned.
        There are plenty of advanced statistical tools and methods for transactional processes, short runs, non-normal measurements, etc.  As for the large data bases – the people involved in data mining and large data base manipulation have developed some very fine techniques for extracting information.  There are short courses out there which attempt to teach these methods.  When dealing with people who have taken these courses and who have no prior knowledge of the basics, it has been my experience that their potential for “going wrong with great assurance”  is very high and their ability to detect that they have done so is very low.
      As many posts to this forum wound confirm, this is not to say the current course content is never misunderstood. (For example – do a search on “central limit” and bean count the number of times individuals have posted the “fact” that if you have more than 30 data points the distribution of these points will approach the normal.) However, my experience suggests it is far easier for an individual to detect and correct the mistakes they make with the basic tools.
      Black belt statistical training is the absolute bare bones minimum. To view it as anything else is folly.  I view BB statistical training as an attempt to provide an individual some kind of sound understanding of basic (but useful) statistical methods with the understanding that the individual will be able to use these basics to solve problems and to judge and understand the value of advanced methods as they are offered as tools for problem solving.

    0
    #118674

    Mike Carnell
    Participant

    Six Sigma Tom,
    I have no intention of getting involved in your discussion beyond the point of making sure you understand that 2 of your 3 points are flawed or completely inaccurate. Robert Butler did a good job, as usual, with the stats side.
    “They were designed in an era when data were scarce and computing power was expensive. We now have gigantic data warehouses and everybody has a super computer on their desk.” At Motorola there was virtualy no issue with data availability. The shock for most of us coming from Motorola to consulting  on Six Sigma was and still is the lack of data in almost every aspect of business particularly transaction and service type projects.
    “They were designed for mass production manufacturing processes. We now have transactional processes, high variety production, short runs, etc.” A fair amount of the early work on DMAIC was at the Government Electronics Sector in Phoenix, Arizona. The programs were predominately low volume. In shared services areas i.e.central wave solder areas, cabling, parts prep, etc, we ran low volume high mix. Mario Perez Wilson’s book “Six Sigma” is about the FMU 139 bomb fuse line which was our highest volume line and that was less than 700 per day – medium volume at best.
    Good luck.

    0
    #118675

    BTDT
    Participant

    SST:
    The backbone of all statistical tests is to distinguish whether an effect is random or not random. This has never changed in the history of the discipline.
    It is not possible or necessary to give you anything beyond an introductory course in statistical analysis during training. If this were not the case, then the ideal BB would be a statistician – clearly not true.
    It is usually not necessary to get much more involved in advanced techniques when there is so much more involved in executing a project than statistical analysis. I have seen too many good transactional projects with great savings executed with only the seven basic tools with no p-values to think BB training requires more statistical tools.
    Most new work done in statistics is developing new techniques, usually to deal with an unreliable or incomplete data set. These new tests are named after their developers and are highly specialized. It is amusing to talk to a crowd of statistics gurus and see them split into two groups depending on whether they are the ‘objective frequentists’ or the ‘subjective Bayesians’. Both approaches will give you the same result, but the groups will still square off and throw buns at each other during the conference dinners. I have used and developed techniques based on both philosophies depending on the problem at hand.
    The only reason I could get away with doing a Bayesian Markov Chain Monte Carlo determination of the parameters of overlapping processes in an accounts receivable project was that the CEO had a PhD in nuclear physics and used the same technique in his thesis project while designing refueling schedules for nuclear reactors. I read his thesis before giving the pitch. None of the Six Sigma crowd objected because the project determined when was the best time to call the customers about outstanding accounts (and the CEO liked it too). The Six Sigma culture was mature and receptive.
    One of the worst projects I’m aware of was attempting a redesign of an aluminum smelting operation. It was a tremendously complex problem that has resisted improvement for decades. The outside consultants pitched it as a pilot project early in the Six Sigma implementation and suggested a similarly complex DOE to tackle the problem. The report outs were like science fair projects. The corporate culture was not right and the project failed.
    I’m sure others would agree that 90% of their transactional projects could be executed with fewer statistical tools rather than more.
    The ideal BB training for transactional projects? Spend LESS time on stats and MORE time on change management.
    BTDT

    0
    #118676

    Dayton
    Member

    I think that you have missed the point.   Six Sigma does not require the use of specific statistical tools it requires the use of the appropriate statistical tools for the issue at hand.   Some Six Sigma practitioners better understand the selection and use of statistical tools than do others.  
     
    Which statistical formulas and methods are you touting as new and advanced?  
     
    It’s my belief that mathematics and statistics has not evolved as a field as much as has the availability and the ready use of flexible programs such as SAS, SPSS, Statgraphics, Minitab, etc., on office and home PC’s.  It’s also much more the case now that one does not have to be a statistician to effectively use and understand the use of statistics. 
     
    Since you’ve thrown out a provocative thought like this, precisely which statistical tools and methods are you referring to?   What makes you believe that statistically proficient Six Sigma practitioners are not availing themselves of the appropriate tools?  Or are you suggesting that Six Sigma practitioners and mathematicians/statisticians are mutually exclusive?  If so where is your data?  If not, where is your argument?  
    Vinny

    0
    #118677

    Six Sigma Tom
    Member

    My point is that effective change management and leadership requires that we identify root causes and create a proper transfer function linking these causes to the Y we’re trying to change. Outdated statistical tools make this more difficult.Statistical tools are more than a “starting point”, they’re a core element of Six Sigma. Take statistics out of the mix and see what you have left, it won’t be half as successful as Six Sigma. It’s like science or engineering without the math.Since you mentioned it, here’s something new to throw in your toolbox: the method of multiple hypotheses. This approach requires that you formulate several alternative models of cause and effect, then use data to choose the model that best fits. It’s much more than stepwise regression, which just crunches different linear regressions on the same set of Xs. It’s a replacement for the usual approach where we state and test an irrelevant and uninteresting null hypothesis (which, by the way, is never true.)Many eminent statisticans have avocated that we abandon the traditional approach to statistical hypothesis testing, and many statisticians and researchers have. But most Six Sigma practioners know nothing else.

    0
    #118678

    Six Sigma Tom
    Member

    Mike,Sorry you won’t get involved in this discussion, I thought you’d find it interesting. Your thoughtful responses are always interesting and welcome.Anyway, I’m aware that despite the ocean of data we often can’t find the drinking glass of information we need to make improvements. Most data are collected for purposes unrelated to performance improvement, such as scheduling, billing, legal requirements, etc. It’s difficult to use this to establish cause and effect. On the other hand, these data ARE often (but not always) useful for ASKING QUESTIONS.When I’m talking about methods developed in an era of scarce data I’m talking the period from 1919 through 1949. This is when control charts, DOE, and most other statistical methods used in Six Sigma were developed. Most statistical algorithms coded into computers were developed before 1980, when computer power was more expensive. Six Sigma hasn’t added much (if anything) in the areas of innovative new statistical tools or powerful data crunching applications. That’s not what it was designed to do. However, I think Six Sigma would be improved if we shamelessly borrowed some of the innovations that have been developed in these areas in other fields. There are a lot of them.I don’t think it makes sense for Six Sigma practioners to be resistant to change. Isn’t improvement (change) what we’re all about?

    0
    #118679

    Six Sigma Tom
    Member

    One new tool that I’ve found to be useful is the classification tree. This involves creating a tree where the X having the greatest impact on the Y is shown at the top and “split” in a manner that maximizes Chi-square. Then each X split is examined and the X that has the most impact on it is found and split. The process continues until some stopping rule is reached. The result is a “Tree” that shows a statistical model that explain the top level Y. It’s easy to explain to others, automated, and often leads to insights that help subject matter experts develop models. The Black Belt can then develop DOEs or other tests of these models. I’ve seen a lot of improvement driven by this approach.SPSS has a product called AnswerTree that does this analysis. You can see it at http://www.spss.com/Free/atree.htm.

    0
    #118680

    Six Sigma Tom
    Member

    I see that I’m in a huge minority here. Am I the only one who thinks this is the case? Has no one else looked at a control chart with 10000 plotted points and wondered what the #$%& it told them? Has no one else tried to explain P-values and null hypothesis testing to a process owner and received a blank stare in return? Or looked at a database with 1,000,000 records and wondered how to use classical tools to find the nuggets of gold that were surely buried in there?Six Sigma Tom

    0
    #118681

    Robert Butler
    Participant

      “Has no one else looked at a control chart with 10000 plotted points and wondered what the #$%& it told them? Has no one else tried to explain P-values and null hypothesis testing to a process owner and received a blank stare in return? Or looked at a database with 1,000,000 records and wondered how to use classical tools to find the nuggets of gold that were surely buried in there?”
      Yes to all of the above.  When I’m confronted with a “control chart” with 10,000 plus data points the first thing I ask is how was it built.  The usual answer is “we have software that samples the product and plots it for us.”  ….and what does that tell me – it tells me everyone involved hasn’t a clue as to what a control chart is or what it is supposed to do. It also tells me there will probably be severe autocorrelation present and because of this the people using this data for “control” have been making a passle of unneeded changes to their system. What do I do with it – it depends.  If after questioning I decide the data is representative I get the data file and run it through a time series analysis in order to identify the separation in samples needed in order to treat them as independent.  I then take the much smaller data set and put it through the usual wringer for assessment of process behavior.  What does the customer get out of it – typically a drastic reduction in needed testing and sampling, a huge drop in adjustments to the process, and a much more uniform product.
      P values and hypothesis testing – usually I don’t even bother them with such things.  I put all of this in the following form:
      “I have found the following things, differences, changes, etc. to be statistically significant. Question: do any of these things differences, changes matter to you physically?”  If they do then we go to the next step of investigation – If they don’t – well, I’m not married to them either so we leave them and sit down and give the problem some more thought.
     Data sets of a 1,000,000 records – first question is how was this put together.  If it is someone’s idea of a meta analysis – I stop right there and we get down to finding out just how “similar” all of these data sets really are.  If it is a genuine block of 1,000,000 consistent records then the questions shift to that of data structure – missing data, correlation structure, etc.  In short, huge data sets require a lot of non-statistical checking before you even think of trying to analyze them. If the structure of the data is acceptable and if there is really signal present in the data you will find the usual tools of statistics are quite adept at finding them.
      I realize some of the above could come across as sarcastic and perhaps even flippant – I don’t mean it to be.  I also don’t want you to think I believe all of statistics can be reduced to a few simple things – it can’t – however, the problems you have described fall more into the arena of defining, measuring, and explaining to the customer as opposed to the area of statistical analysis (of which, because of my occupation, I happen to be very fond).

    0
    #118682

    BTDT
    Participant

    SST:I have used Amos, Clementine, and beta tested MineSet(SGI). Neural nets all have equivalent techniques based in well understood, traditional regression techniques, and lack the statistical rigor associated with a good Analyze phase.
    We have, however, incorporated some of these components into the fraud detection component of the Control phase of a transactional processing application. I would hesitate to use them in general to replace existing, simple, rigorous and accepted tools.
    I see no value in adding data mining tools to the BB toolkit as part of the Analyze phase. It is difficult to present the results of the analyses to even the statistically oriented Six Sigma crowd. I have only touched on more advanced tools and even then it was for experienced, motivated MBBs.
    The ‘nuggets’ you speak of sound suspiciously like the marketing jargon used by the data miners.
    The error in sampling does not really change the result whether you are sampling 1,000 points or 1,000,000. I would question the business impact of a result that would require a sample as large as that to demonstrate a significant difference.
    Robert Butler has good points about the amount of work that can be done in a 2 week section of BB training. If I can get BBs to apply the simpler statistical tools correctly, then I can go home happy.
    BTDT

    0
    #118686

    Dayton
    Member

    You might not be in the minority had we known what your concerns really were. Why didn’t you just start by asking something like does anyone see the need for or have they used neural networks versus decision trees (and are you selling CART or CHAID tools?) or has the absence of effective data mining techniques and algorithms limited their data gathering or analysis for Six Sigma projects?   Are you trying to parse out a concern regarding the distinction between analytical exploration and exploitation?   One is the finding of new data and the other the better understanding of what’s already in front of you.
    If you were trying to elicit responses regarding data mining you could have started there.  You started by describing, it appeared, your concerns regarding the limited use of statistics or mathematics by Six Sigma practitioners and then moved into the benefits of using data mining to locate, gather, sort, analyze and better profit from the data pearls in the ocean.
    I personally would not be reluctant at all to bring the discussion of data mining applications and techniques into the Six Sigma domain and feel that there are certainly analytical circumstances in which data mining can play a significant role.  But I would not start off by saying that Six Sigma per se lacks mathematical or statistical rigor. 
    Since you seem to want to center on data mining where would you like to begin?   Areas that can be interesting fodder for discussion could be decision trees, genetic algorithms, rule induction, artificial neural networks, or the nearest neighbor method.  Or, as has already been implied by other posters, are you selling data mining tools or consulting services?Vinny

    0
    #118697

    Anonymous
    Participant

    Vinny:
    What is the difference between CART (Classification & Regression Trees) and CHAID (Chi-Square Automatic Interaction Detection)?  I recall using these tools many years ago in a package called KnowledgeSeeker.
    I noticed that JMP now has Recursive Partitioning.  Is this the same thing?
    What about the dangers of such tools?  Gunter wrote an article in 1996, Quality Progress pointing out the dangers od data mining called “Data Mining: Mother Lode or Fool’s Gold”
    Another very similar article called “Data mining for fool’s gold”
    http://www.cba.ua.edu/~mhardin/Data%20mining%20for%20fool’s%20gold%20-%20Computerworld.htm
     

    0
    #118698

    Dayton
    Member

    Good question.   You always hear one spoken in the same breath as the other just as though they are either one in the same or so uniquely commonly directional in their use that the two decision tree approaches yield the same endpoint although their mathematical and graphical structures are somewhat different.   As I understand it, CART’s benefits are primarily in producing very abbreviated or reduced binary decision rule sets due to having tight tree branch pruning defaults and CHAID utilizes multiple splits instead of the binary splits used in CART.  That’s greatly oversimplifying it but I believe represents the real essence of the applied nature differences. 
     
    Many of the commercially available data mining approaches use various elements of the several topics I noted in my response to the initial poster.   From what I’ve seen the uniqueness comes from combinatorial approach and sequencing and a somewhat situationally appropriate direction – that’s where the smart people in the field reside, in that space beyond knowing when and how but also the applied why.
     
    The danger of using these or any tools is in becoming so enamored with the mechanics of the program and the results provided that you lose sight of the fact that they are really just analytical aids and not answer providers.   You need to fully understand what you are asking and the context of results provided – they don’t take the place of analysis they can aid you in analysis.   Plus anytime you are using decision tree analysis of any type incorrect data collection and use upstream can land you in a bad place downstream. 
     
    Where were you heading in your questions?   What have been your experiences in using various data mining methodologies and what do you perceive to be the differences between CART and CHAID?   What has been your experience with the use of neural networks and has that been dependent on node placement, selection and algorithm development?  Have you developed data mining algorithms and what were the applied results of your work?
     
    I had the great pleasure to have worked with Dr. Kurt Thearling in the past on a post-doc expansion and if there is a person out there with the capabilities of going well past third generation data mining and neural networks into enhanced self-generating artificial intelligence it’s Kurt. 
     
    What do you know about these topics, how did you learn it and what have you done with it?  And, most importantly, what’s next and where do you go from here?  
     Vinny

    0
    #118699

    Six Sigma Tom
    Member

    First, I’m not selling anything.Second, I’m not only interested in data mining, although I do think it’s a useful tool that we should at least make the Belts aware of.I stand by the premise of my forum topic: Six Sigma Stats are old fashioned and flawed.1. The statistical methods taught to Six Sigma black belts are old fashioned. 99% of what we teach belts are traditional methods that have been around for 50+ years. True, we can’t possibly expect them to learn everything that’s new, but it would be worthwhile for the six sigma community to ask which new techniques should be taught.2. The statistical methods taught to Six Sigma black belts are flawed. The scientific and professional statistician communities have reached a consensus that the traditional statistical methods taught in universities are logically flawed. However, they’re struggling to get new methods introduced in universities. The Six Sigma community could take the lead by changing how we teach statistics to Belts.I must admit that I am surprised at the resistance I’ve encountered in this forum. I would’ve thought change agents such as we would be more open to the suggestion that perhaps an imporant aspect of Six Sigma might be improved, thereby improving our results. Perhaps you’re correct that I went about it wrong. I admit that my choice of words is provocative and (apparently) inflamatory. But then, I’m a Six Sigma guy, not a diplomat. ;-).Six Sigma Tom

    0
    #118700

    Dayton
    Member

    No.  I’m not trying to be overly diplomatic either but, on the other hand, certainly don’t want to appear inflamed – as I’m not.  I simply asked you what you meant by:
     
    The statistical methods taught to Six Sigma black belts are flawed. The scientific and professional statistician communities have reached a consensus that the traditional statistical methods taught in universities are logically flawed. However, they’re struggling to get new methods introduced in universities. The Six Sigma community could take the lead by changing how we teach statistics to Belts.
     
    But, I still have not heard example.   Surely you know that there is not a lot of new statistics or math out there, there are derivations of centuries old stuff and most of the new comes in automated calculation and more effective use of the same old stuff.  What are you talking about?   Get to the point or stop blabbering about the lack of innovatively applied “new” statistical tools out of the Six Sigma community.
     
    You have yet to convince me that you are either Six Sigma trained or that you know what you are talking about in regard to statistics or mathematics.   So far you are regurgitating things that you claim to have read or what you believe to be the position of informed academics.  Change my mind, change agent.  :-(  
     Vinny

    0
    #118701

    Ken Feldman
    Participant

    I have patiently followed this thread looking for the great revelation from you as to specific examples.  Let’s see if we can get some direct responses:
    1.  What tool do you propose that is better than control charts for monitoring process change over time?  Why?
    2.  What tool do you propose that is better than a two sample t test for distinguishing whether the mean of some continuous variable process has changed from some original state? Why?
    3.  What tool do you propose that is better than multiple regression for establishing a correlation and prediction for a continuous Y variable and multiple continuous X variables?  Why?
    Please answer these directly or join Mr. Annoying and Reigle in the time out corner.  I look forward to your specific response, thanks.

    0
    #118704

    New VGB
    Participant

    Congratulation for  your  “new”suggestion.Please go forward and  don’t let anybody furstrate you……..

    0
    #118707

    BTDT
    Participant

    SST:Let us do a bit of DFSS and change management.
    Part of being a successful BB is to be a diplomat of some sort. As a change agent yourself, recognize the first step in getting a project off the ground is to understand and acknowledge if the stakeholders have a problem.
    1) Accounting is old fashioned. I trust it. What is the 1% we teach that is not old? Fisher’s exact test is superior to the Chi square, but was only recently included in software because it requires calculating factorials up to the number in the sample size.
    2) What is flawed about the tests we use? Evaluate a t-test with a bootstrap (recent technique) and you’ll find it is pretty robust to the assumptions. I see no evidence that the traditional methods taught in universities are logically flawed.
    Surely not logic alone is the criteria. There is nothing that seems more illogical than a screening design in a DOE, but it works.
    One of the first questions I ask a class is how many people took a stats course at the college level. Most hands go up (including mine). I then ask how many people really understood it (most hands come down, including my own). The biggest problem in teaching the existing methods to either a BB class or university group has nothing to do with the analytical tests.
    If you have tools you wish to use to augment the ‘body of knowledge’, then present it to your own BBs and check for resistance. Have you ever faced a group of HR people and told them about DOE? If you are successful, then tell us all about it. Nothings sells like a good example with endorsements.
    All the best,
    BTDT

    0
    #118709

    Royale with Cheese
    Member

    Oooo…checkmate.

    0
    #118710

    Dayton
    Member

    Another fascinating rising to the bait floated overhead.   You came out of the water like a fat trout seeing one of Ken Iwamasa’s Small Western Green Drake flies break the water plane.  
     
    Where do you find significance in what was said about the lack of statistical rigor and appropriateness of same as used by Six Sigma practitioners?  Telling someone who is trying to engage topical debate that their “new” suggestion has merit prior to them doing other than attracting attention saying that they have a new suggestion and in advance of them actually offering up a new suggestion is really just adapting an antagonistic role versus a contributory role on your part.  Don’t you think?  
     
    I hope that SST was encouraged by your acceptance of and alignment with his thoughts.  The one or two times that you have agreed with me gave me cause to go back and rethink my expressed position or opinion.   But like you and Six Sigma Tom, sometimes I just like to take a position and argue it out to see where it goes. 
     
    Vinny

    0
    #118711

    Mikel
    Member

    when and how were hypothesis tests discredited?
     

    0
    #118712

    Mikel
    Member

    the tools he has referenced so far are in Shainin’s toolbag – want to talk about out of date?

    0
    #118713

    Anonymous
    Participant

    Thanks for your answers.  I used CHAID and CART back in the late 80’s simply as an alternative to traditional ANOVA/Regression.  (The reason that I asked the question is that I did not know what the differences were. The KnowledgeSeeker tool incorporated both approaches).  We had some good success with that tool, but used it to give us insight into factors that would go into a DOE.  Hence the reference to Gunter’s article.
    Your point about being an aid to analysis is well taken.
    My use of Neural Nets has been limited to predictive modelling. I developed a model (using Neuralware Predict) for a mining company that provided a 24 hour forecast to schedule train requirements to haul ore.  I started with classical regression, also used harmonic regression, time series analysis, regression on principal compenents and finally went to the Neural Net. It cut my prediction errors in half and consistently did so.  The challenge was that the model needed tuning on a monthly basis.

    0
    #118714

    Dayton
    Member

    My use of Neural Nets has been limited to predictive modeling. I developed a model (using Neuralware Predict) for a mining company that provided a 24 hour forecast to schedule train requirements to haul ore.  I started with classical regression, also used harmonic regression, time series analysis, regression on principal components and finally went to the Neural Net.
     
    Still some pretty cool stuff – it was cutting edge back then.   Do you still play?    I have a couple of friends with Santa Fe Institute that are currently using neural networks (probably nth generation nets) in various interdisciplinary complexity studies
     Vinny

    0
    #118715

    Anonymous
    Participant

    SST,
    Why don’t you get specific.  Let’s start with the basics. Tell us how a One Sample t-Test is logically flawed.  Are you going to recommend Equivalence Testing?
     
     

    0
    #118716

    Anonymous
    Participant

    Sorry it’s been a while since I’ve played. Sounds like I should get back into it!

    0
    #118717

    Six Sigma Tom
    Member

    1 sample t tests begin with by stating a null hypothesis about the mean, i.e., mu = ###. Without taking a single sample, I can tell you that the null hypothesis is false. Is this not a logical flaw?

    0
    #118718

    Six Sigma Tom
    Member

    1. What tool do you propose that is better than control charts for monitoring process change over time? Why?Personally, I love control charts. They’re nice, graphical tools that are easy to understand. One of my favorites. However, they’re often used for things other than monitoring process change in real time. Such as plotting vast amounts of historical data. For this “data mining” purpose, they suck.2. What tool do you propose that is better than a two sample t test for distinguishing whether the mean of some continuous variable process has changed from some original state? Why?The mean has changed from its original state. You don’t need a t-test to tell you that.3. What tool do you propose that is better than multiple regression for establishing a correlation and prediction for a continuous Y variable and multiple continuous X variables? Why?Regression is a powerful tool. We nearly always focus BB training entirely on linear regression. What about all of the other variations of regression (e.g., non-linear, ridge regression)? What about graphs and charts of the relationships? What about resampling/bootstrapping to perform regressions and create models that don’t require as many assumptions?Darth, are you really so in love with the status-quo that you can’t see any way to improve?

    0
    #118719

    Six Sigma Tom
    Member

    1) Accounting is old fashioned. I trust it.
    Accounting is a farce. It’s as trustworthy as Enron and the social security trust fund.2) …I see no evidence that the traditional methods taught in universities are logically flawed.Read Deming’s writings about enumerative and analytical statistics and see if you still feel this way.”Surely not logic alone is the criteria. There is nothing that seems more illogical than a screening design in a DOE, but it works.”I find that sequential DOEs, starting with a screening design, are eminently logical. I find the use of enumerative statistical methods to address analytical statistical problems and statistical hypothesis inference testing to be illogical and impractical. They often don’t work.In fact, I have faced HR people and told them about DOE. They find it exciting. I’ve seen HR BBs do great DOEs, such as using DOEs to evaluate different methods of selecting, training, and coaching sales employees.

    0
    #118720

    Six Sigma Tom
    Member

    Stan,Here are a few references in no particular order to get you started:Tukey, John W., “Analyzing data: sanctification or detective work?” American Psychologist, unfortunately my copy doesn’t show the date. (includes an appendix on the Jackknife, a clever approach we’d do well to use more often)”Loftus, Geoffrey R., “A picture is worht a thousand p-values: on the irrelevance of hypothesis testing in the microcomputer age,” Behaviour Research Methods, Instruments, and Computers, 1993, 25(2), 250-256.Cohen, Jacob, “The earth is round (p < .05)," American Psychologist, December 1994, 997-1003.
    Wang, Chamont, Sense and nonsense of statistical inference, New York: Marcel-Dekker, 1993. (Book)Dixon, Peter and O’Reilly, Tenaha, “Scientific versus statistical inference.” Canadian Journal of Experimental Psychology, 1999, 53:2, 133-149.”The method of multiple working hypotheses,” Science editorial, Vol XV, No. 366, February 7, 1890.Six Sigma Tom

    0
    #118721

    Six Sigma Tom
    Member

    By the way, what is equivalence testing? Do you think it might be useful in performance improvement?

    0
    #118722

    Six Sigma Tom
    Member

    You have yet to convince me that you are either Six Sigma trained or that you know what you are talking about in regard to statistics or mathematics.Ah, ad hominem attack. I wondered how long it would take for this to appear.But, I still have not heard example. Surely you know that there is not a lot of new statistics or math out there, there are derivations of centuries old stuff and most of the new comes in automated calculation and more effective use of the same old stuff.Aren’t “automated calculation and more effective use of the same old stuff” new? That’s like saying that computers are nothing new because they just let us do things faster. Maybe not “new” in the philosophical sense, but jeez, let’s at least see if we can use some of this old-but-faster-and-more-effective stuff!

    0
    #118727

    EdG
    Participant

     Yea Darth, quit living with the status-quo.  Face it, statistics is dead.  Just as dead as the Sith!  (Oops, hope I didn’t give away what happens in book VI.)  No more regression modeling or hypothesis testing, we don’t need proof.  Just look at the data and you can tell.  Close your eyes, take a deep breath, and become one with the data.  Use the Force…
    Hey, Stan or Vinny.  Can one of you alert the shareholders of MiniTAB that they had better bail (or find a new product)?!?!?!
    Now, leave me alone.  I am meditating.  I wish to become more in touch with my data. 
    Shush…

    0
    #118729

    Kraft
    Participant

    I couldn’t agree more, the statistics behind Six Sigma are out dated and no longer effective. Take up arms to banish them from your operations.
    ……please forward your client list. We’re always looking to expand our operations, it might as well be the business you currently have.
     
     

    0
    #118730

    Robert Butler
    Participant

      The purpose of a t-test is not to check if there is a numerical difference between two averages – the purpose of a t-test, or any of the other hypothesis tests for that matter, is to check to see if the observed difference matters at some predetermined acceptable level of certainty.
      Linear regression does not mean a straight line as opposed to a curved line. Linear regression means the model is linear in the parameters – there are lots of linear models whose plot is anything but a straight line – check Daniel and Wood Fitting Equations to Data 2nd edition pp.20 -23 for example. 
      Non-linear regression (as opposed to non-linear black box regression) means non-linear in the parameters and typically requires a mathematical expression of the problem which has been derived from first principles.  This model is necessary because you usually have to take its derivatives and express these to the machine along with first guess estimates of the parameters of interest.  While this can be done the biggest drawback is the need for an initial model – something which, more often than not, is simply not available.
      The tone of the posts suggests it is the last class – non-linear black box regression – that may be the focus of the questions and the interest. Non-linear black box regression methods like MARS and neural nets are mapping methods in the sense that they connect inputs to outputs. If you have absolutely no interest in what governs how things map (the cited example of credit card fraud is a good one – another one, which I worked on and helped bring to market, was that of determining the amount of fluid in a container whose geometric shape was so contorted as to make any other kind of measurements almost impossible) these can be useful tools.  The problem with these methods is if you have to explain X-Y relationships from the standpoint of possible cause and effect you won’t be able to do this in a manner that is going to satisfy most people. 
        Since plausible explanations of the relationships between X and Y are central to most BB efforts the simple mapping of inputs to outputs – no matter how well the process can be mapped and/or controlled with MARS or a net – precludes their utility.
      I don’t understand the question “What about resampling/bootstrapping to perform regressions and create models that don’t require as many assumptions?”  How many assumptions are we talking about?  When you run a regression all you are doing is making a number of algebraic calculations based on the simple principles of least squares – nothing more. 
      When you go to test the model you make the following three assumptions:
    1. The error is a random variable with mean zero and unknown variance sigma squared.
    2. Yi and Yj are uncorrelated
    3. The error is a normally distributed random variable.
      This last implies the error terms are not only uncorrelated but necessarily independent (Draper and Smith pp.22-23 Applied Regression Analysis, 2nd edition). Regular linear regression has dozens of tools to help in the instances when there are issues with #2 and #3.
      As for graphs and charts of relationships – no problem – this is done (or should be done) all the time.  Tufte – The Visual Display of Quantitative Information, Visual Explanations, and Envisioning Inforation describes (or rather illustrates) all kinds of ways to do this and if you have a decent statistics package you have access to many such visual tools.
      Do I have a problem teaching things line MARS and nets to BB’s or anyone else – no – but at the end of the day BB’s are going to want to know the utility of these methods. For most of the things they will be asked to do my experience suggests their time would be better spent learning something else.

    0
    #118731

    FTSBB
    Participant

    Sounds like the problem isn’t the tools, but people’s use of the tools.  That doesn’t mean the tool is flawed, just the execution.

    0
    #118732

    Dayton
    Member

    No, I leveled neither an ad hominem nor a circumstantially ad hominem attack.   I did not attack you I was describing the fact that you were making programmatic assertions of deficiency without your first having laid ground work of competency.   If that is your modus operandi it is what it is.   It seems to fit in well with the forum.
     
    If, on the other hand, you really want to engage in productive debate you might try describing your target, aiming and firing before asking for commentary regarding your marksmanship.  
     
    By now saying:
     
    Aren’t “automated calculation and more effective use of the same old stuff” new? That’s like saying that computers are nothing new because they just let us do things faster. Maybe not “new” in the philosophical sense, but jeez, let’s at least see if we can use some of this old-but-faster-and-more-effective stuff!
     
    You have capitulated regarding your initial thrust of:
     
    I stand by the premise of my forum topic: Six Sigma Stats are old fashioned and flawed.
    1. The statistical methods taught to Six Sigma black belts are old fashioned. 99% of what we teach belts are traditional methods that have been around for 50+ years. True, we can’t possibly expect them to learn everything that’s new, but it would be worthwhile for the six sigma community to ask which new techniques should be taught.
    2. The statistical methods taught to Six Sigma black belts are flawed. The scientific and professional statistician communities have reached a consensus that the traditional statistical methods taught in universities are logically flawed. However, they’re struggling to get new methods introduced in universities. The Six Sigma community could take the lead by changing how we teach statistics to Belts.
     
    Which satisfies me as having won our discussion debate and correspondingly your having lost the discussion debate – on a points-awarded basis, of course.   You need to better make your point and stand your ground.   
     Vinny

    0
    #118736

    Ken Feldman
    Participant

    Unfortunately Tom, you refuted your own argument.
    1.  You presented no option nor any flaw with control charts.  They are as relevant today, if used properly, as they were when Shewhart developed them.
    2.  You are either an idiot or don’t understand the difference between statistical significance as it pertains to change in the mean and change due to random sampling.  Again, no option or flaw.
    3.  Don’t know where you are teaching BB but other types of regression are taught besides linear.  Response surface modeling is usually taught in BB training along with quadratic, cubic and everyone’s favorite, Logistic.  Again, where is the flaw.  What is the specific option you are presenting?
    If you present a logical flawless alternative to the status-quo you might find some interest.  But, you haven’t presented a valid argument yet.

    0
    #118739

    Ken Feldman
    Participant

    Hey Ed, if I was comfortable with the status quo I’d still be working back at the old place. 
    BTW, I spoke with Lukas and I understand that this is not the last Episode of SW. There will be one more where they introduce someone even more evil than Darth Vader……Darth Stans, a multi personality villian.   Plus Jar Jar is being replaced with another character who makes even less sense when he rambles, Vin Vin.  C3PO is being replaced by someone even more prissy and stitled in speech, Reigle69.  Princess Leia is replaced by Babe of a Sister.  Carnookie replaces the plain old Wookie.  Dog Sxxt becomes the loveable R2D2 as the thing you love to hate who rarely does more than make random sounds.  The sauve and silver tongued Phil becomes Han Solo but replaces the Millenium Falcon with his Millenium Mercury.  The revered Yoda, Obi Wan and Luke will continue to be themselves since I owe a lot to them at both a professional and personal level.  Emperor Cyger becomes the evil and controlling WebMaster who seeks to control the Universe and can make a Poster vanish into Cyger Space.  Please feel free to add characters unless you find this whole idea Old Fashioned and Flawed.  Beats the heck out of the current thread since the little Ewok Tom never made sense anyway.

    0
    #118740

    Anonymous
    Guest

    FTSBB,
    I agree with your conclusion – this is by far the most serious problem associated with the application of all statistical methods.
    Good call ..
    Andy

    0
    #118741

    Dayton
    Member

    Plus Jar Jar is being replaced with another character who makes even less sense when he rambles, Vin Vin. 
     
    OK, now you’ve done it.   It’s not my fault your cranial receptor decoder patch slipped off the bald spot as your oversized helmet slapped around while you chased Tom Tom down the path of limited statistical rigor.    I can’t slow down and explain everything to you and as a result it just seems like gibberish to you.    Well maybe I can, just this once…   Ggeee Ddaaarrtthh, Iiii wwaasss oonnllyyy ssspppeeeaakkkiinnggg mmeeetttaaappphhhooorrriiiiccaallllyyyy…  Ssooorrrrrryyyy tttoooo hhaavveee cccooonnnnfffuussseeddd yyyooouuuuu……
     
    Well, that wasn’t so bad after all – a little time consuming but worth it for assuring clarity. 
     Vinny

    0
    #118742

    Ken Feldman
    Participant

    Better, thanks.  I assume it was your GiantMedicalProducts, Inc. that provided all the prosthetics I used in the new film.  I am glad that there are companies like yours to allow me to live a full and evil life despite my unfortunate accident.  Just think, no medical products, no Darth, no BILLION DOLLAR payoff for Lukas and his gang.  Then I would have to post under my real name, Shirley.

    0
    #118744

    Ken Feldman
    Participant

    Better, thanks.  I assume it was your GiantMedicalProducts, Inc. that provided all the prosthetics I used in the new film.  I am glad that there are companies like yours to allow me to live a full and evil life despite my unfortunate accident.  Just think, no medical products, no Darth, no BILLION DOLLAR payoff for Lukas and his gang.  Then I would have to post under my real name, Shirley.

    0
    #118745

    EdG
    Participant

    Shush, enough rambling.  I am still staring at my computer screen and trying to become one with my data.   Hope I begin to sense the Force soon.  I am starting to get a migraine…
    So, who does Mikel Harry get to play?  Luke’s Uncle Owen?  (They kind of look alike…)
     

    0
    #118746

    Anonymous
    Participant

    Interesting point about H0, however in practice the classical tools work very well.
    For your information, equivalence testing was developed in biostatistics to establish the equal effectiveness of two drugs (say brand name versus generic).
    In equivalence testing, the null hypothesis is that the two groups are not equivalent to one another, and hence rejection of the null indicates that the two groups are equivalent.  In essence you are using two one-sided t-tests.
    See for example:
    http://www.fda.gov/ohrms/dockets/ac/01/slides/3735s1_02_Lachenbruch/
    http://www.mors.org/meetings/test_eval/presentations/C_Warner.pdf

    0
    #118748

    BTDT
    Participant

    SST:
    You have still not convinced me that I have a problem that requires fixing. If you want the process owner (me) to change, then leading me will work better than pushing me.
    The Royal Statistical Society doesn’t agree with your points about statistics being flawed. They have a very clear idea about where the various tools fall in the context of other techniques.
    http://www.rss.org.uk/main.asp?page=1802
    Please contact them if you feel the Graduate Diploma program in ‘Statistics for Industry and Quality Improvment’ requires changes.
    You may feel I am resistant to change, but I am right 19 times out of 20. I see no immediate need to abandon what is working quite well.
    Regards, BTDT

    0
    #118752

    luke skywalker
    Participant

    This has been a fun thread to read, as I sit in the back of a class of new MBBCs. I honestly have to say, though, I still can’t tell what you’re truly seeking here. Maybe to prod the pros to gain a sense of entrenchment of dogma, or maybe you’re just bored and wanted to wind up the regulars. Either way, we’ve seen some decent points on how tools and concepts have evolved over time.
    Without debating tool for tool, it seems that critical or rational thought is the “tool” that we need to “teach” most effectively. Everything else is just that – a tool, which serves to help us solve our problems more effectively.
    I almost get the sense that you were roasted in a discussion with a stat-magician who trashed our approach. Maybe I’m off in my read of the Force.
    In any event, I agree that while mainstream SS uses ‘classical’ stats, there are planty of areas where much newer methods are employed and computing systems are leveraged to complete complex algorithms. So what would you suggest that we as a community do to improve our position to get past the ‘flaws’ you are holding dearly to?
    Oh, and thanks for the props, Pappy.

    0
    #118765

    JO Club CEO
    Participant

    Six Sigma Tom – we love you and admire you, what a breath of fresh air out here on the forum when we are so used to Stan, that Reigle dude, and lest we forget the infamous bore who answers each and every post…..Darth.

    0
    #118766

    Markert
    Participant

    Darth,
    Another fine job in the imagination department, masked man. I am happy that again, I am seen playing the role of handsome goodguy, hoping for extra takes in that extended romantic portion with the lovely Babe of Sister during the new proposed episode.
    If these guys from your South Beach possee are involved in the film, please count me out.
                               
    It appears to be some type of intergalactic hazing ritual involving a pickle….
    P.S. I move that Six Sigma Tom now should be known as Three Sigma Tom. Any second?

    0
    #118769

    Markert
    Participant

    JO;
    Interesting observation. I’d have to admit that Darth has exhibited amazing restraint in his participation for several months. I was concerned a couple of times he forgot about us, but I attribute that to a lot of extra airtravel miles, a challenging and engaging new gig, and plenty of extra trips to the Non-BoA ATM depositing those fat, juicy paychecks and extra bonus incentives for his hard work.
    What have you been up too? Still riding that seatless mopehead around town with the rest of your boys? If I see you guys again at the store, better watch out.
    But we are having a sale on Britney Spears CDs, press on tatoos, and cold juice boxes for you and your JO Club friends. This stuff is 10% off till Saturday!!! Bring a note from your Mom. And don’t forget her, cuz Mom’s Day is Sunday.

    0
    #118784

    Ken Feldman
    Participant

    Missed Ya JO.  How ya been?  The “bore who answers each and every post”.  I’m hurt.  Here is the bet…..if you count the last 100 posts and I answered “each and every” one of them, I will pay for you and your Club at Phil’s HD sale.  Or better yet….if you count the last 100 independent threads and I answered “each and every” one of them I will arrange to have Britney come to the Club and JO “each and every” member.  Remember data rules, not dogmatic unsupported statements.  Do you accept?

    0
    #118785

    Ken Feldman
    Participant

    Phil,
    First of all, I feel the picture is in bad taste and not worthy of the sophistication exhibited on this Forum……gotcha.  No, the SOBE crowd  will not be participating in the making of the film.  They  refused to remove their thongs and sandels and wear the Storm Trooper outfit.  I did notice that the pickle was LARGE and thus a Kosher one.  Of course, we did need investment capital from the S. Beach Jewish population to include all the special effects.

    0
    #118788

    Ken Feldman
    Participant

    Phil,
    I would like to suggest that you include in your sale the free samples of Prozac I have been able to obtain.  Don’t worry about the expiration date, they don’t mean much.  It might help with JO.
    Funny how BOA was happy to part company with me but got a little anxious when I sought to withdraw the $18M in my combined accounts.  So, I still bank with them and yes, the BOAJOC made it worth my while.

    0
    #118794

    Ken Feldman
    Participant

    Phil, your boss will be here on Friday to tell us all about SS at HD.  We will be measuring any change in nose length as he gives his speech.  The last time we did this for a corporate executive, some guy from BOA if I recall, the delta in nose length was exactly 4.587 inches indicating some pretty big fibs about successful implementation.  I will provide specific data upon completion of Mr. Bob’s visit.  If you are intending to go with him, let me know and maybe we can meet and tour the great sites here in         

    0
    #118812

    Dr T and the ladies
    Participant

    Darth – the JO’s were right about you….what a hypocrite you are! 6 months ago you were pumping BOA’s six sigma program like it was the next GE.  Get a life.

    0
    #118813

    dragon ace
    Participant

    Well, if your rating of BOA on scale of 1 -10 went from 2 to 3, they can count you as a delighted customer and old Milton can sign off that he saw a statistically significant change in sat ratings.

    0
    #118814

    Markert
    Participant

    Frequent visitors to isixsigma will note that Darth didn’t even finish his posting, leaving us at “..and tour the great sites here in” shortly after his forehead hit the enter key sending his post to cyberspace.
    It can be revealed that Darth was enjoying a brief lunch at the time. And it is a good thing I was there to help give the evil guru the Heimlich maneuver, helping to dislodge a portion of a badly chewed, hastely swallowed king sized roast beef and cheddar sandwich between his tightly scheduled advanced classes and the counciling of whining green belts. 
    I’ll bet Vinny just would have suggested, in a unparagraphed 300 word essay, one of GiantMedicalProducts’ anti-gagging implant products. But I jumped in an took action….
    Seeing that I had brought lunch and saved his life, and as he is now going to be indebted to me for life, I find it comforting to post my photo and since the Babe of a Sister has several romantic scenes in out upcoming new movie, I’m hoping she’ll post e-pics for us.
    Here’s me, Darth’s lunch buying, heimlich doing, Babe of Sister kissin, Sxxt Sigma SuperDuper TopSecretMasterRedGarderFanRustBelt:
                                
    It’s me, the wisecracking OvenMitt, who doesn’t have a nose to measure or a mutt named Nardingo. But if any of the tour sites include ones where Nard’s 9′ Chia Head CEO bush are located, pay special attention to the sideburn and ear hair. 
    Nope, I still have to stay here at the store and watch for the JO’s and wrestle the gators in the storm retention pond Friday. Sorry. But do try the ‘pull my finger gag with Mr. N’ he loves that one….

    0
    #118818

    Ken Feldman
    Participant

    Wow, and maybe things change also.  Six months ago, Bush was re-elected by a large majority and now his rating is way down.  Many thought the war in Iraq was the right thing while now many are questioning the actions we took.  Things change and so can opinions as new evidence becomes apparent.  Yes, I thought BOA was doing a good job with implementation but things have changed in the last six months.  Hundreds of SS professionals have been let go.  The group I was with is almost at half strength from six months ago.  It’s not being hypocritical, it is re-evaluating based on new data. 

    0
    #118819

    Ken Feldman
    Participant

    But, according to Poster Tom, we don’t need to test whether there is a difference since any idiot knows that 3 is higher than 2.  Here is an interesting tidbit.  Every month, approximately 32,000 cust. sat surveys are collected across the banking centers.  You might say WOW, that’s a sxxt load.  Now you stratify by about 4,000 banking centers.  That leaves about 8 surveys per month per banking center to make decisions on cust sat movement.  What to do now?  How about doing a rolling 3 month average.  Now you have about 24 surveys per banking center across a quarter with which to make decisions.  Any thoughts on how valuable and timely cust sat decisions can be made?

    0
    #118820

    Ken Feldman
    Participant

    Thanks again for the help today.  My darn dentures slipped and I accidently swallowed that piece of roast beef.  But, I am curious, why did I have to bend over and since when do you use a Kosher pickle to expel food caught in the throat?  And why all the giggling?  BTW, it is whining BBs not GBs.  I did note that the landscape people at the Company installed a large shaped bush outside corporate hq today but it was drapped by sheets so I couldn’t see what it was.  I now assume it was the nine foot chia you mentioned.  I will forward your plea to BOAS and encourage her to reply to you.

    0
    #118821

    Mikel
    Member

    The real question that should haunt you the rest of your life is why did I think that George W. was okay. I can see missing on B of A, but Gumby for president? I would think you would sheer your dreadlocks and never look another man in the eye the rest of your life.

    0
    #118823

    Dayton
    Member

    I’ll bet Vinny just would have suggested, in a unparagraphed 300 word essay, one of GiantMedicalProducts’ anti-gagging implant products…
     
     
    That was harsh, rude, completely wrong, and a deliberate mischaracterization of my probable situationally driven response to seeing someone as esteemed as Dr. Darth gagging himself to death on a wad of semi-masticated cheese beef – a circumstance worthy of elevating the emergency condition threat level to an obvious Code Brownish Yellow.  
     
    Granted I would have stopped, assessed the situation, and used my BlackBerry to access GiantMedicalProducts’ product catalog potentially letting poor Dr. Darth succumb to his gluttonous clog in my attempt to order just the right advanced anti-gagging implant from GiantMedicalProducts Customer Service and then FedEx’ing for immediate delivery said implant to the Taco Cabana drive-thru window in question as I cleared my mind for maximized emergency response, focused my priorities, blocked all distracting minutia and immediately sprung into action authoring a combined Taco Cabana Incident Report and peer reviewed Operations Research Journal Article (both of which I thankfully have templates for on my BlackBerry) emphasizing the effective emergency condition usage of Markov Analysis integrating fully optimized wait and queuing nodes assuring best case logistical contribution.  
     
    But, to the point, your egregious and I believe deliberate misstatement was in regard to my alleged lack of paragraphing.  I paragraph well.  I just don’t always use periods.  Say you’re sorry.
     
    Vinny

    0
    #118824

    Kavita Sahu
    Participant

    This is an academic site used by many civilized people. Please do not convert it into a pornographic comedy site.
    Web administrator should take action

    0
    #118825

    indresh
    Participant

    not able to view the same
    please send the same at [email protected]
    thanks

    0
    #118834

    Ken Feldman
    Participant

    ACADEMIC SITE USED BY CIVILIZED PEOPLE.  Are you kidding?  Have you been reading this site for a while?

    0
    #118837

    Mikel
    Member

    They obviously don’t read anything by me, you, Phil, Vinny, Mr. Annoying Reigle,…….

    0
    #118838

    Markert
    Participant

    Darth:
    It is my belief that Kavita Sahu suffers from the punctuation deficiency highlighted by Vinny in a previous post; for example, using punctuation, the post would read:
    Academic’s ite; used by many civilized;  people…. It makes perfect sense. Anyone using “Dr.” as part of their screen name, and the many theorists who propose ‘boil the ocean’ questions, and answers are our academics. Adam B., our new associate Paul, are among a few whom are very civilized and not prone to act up like some of us. And people. That is where the rest of us fall in, apparently.
    I don’t know about you, but I generally am very academic and civilized. I am posting with my pinky fingers extended, just like I act when having afternoon tea with the Royal family. But there must be some kind of metabolism imbalance associated with sitting in front of the laptop, and reading…just reading. And reading more, and more. By the end of the review session of many posts, Dr. Jekyl has turned into Mr. Hyde and the Pavlovian impulse of drooling, hyperactivity, a bad haircut and a love of blood-sport overwhelms me……..I snap.
    I did enjoy Indresh’s response wondering if someone would e-mail him the information on the ‘p-comedy’ site. That’s good comedy.
    By the way, Black Belts who whine around here automatically become green belts. I thought that was standard practice everywhere. I made an assumption in yesterday’s post.

    0
    #118839

    Markert
    Participant
    #118842

    Dayton
    Member

    No, the misrepresentations continue.  It was obviously a completely different imposter at work with some of those.   I am an exceedingly well-paragraphed imposter and won’t say “uncle” until you say “sorry”.   
     
    Besides one of those was from someone called “Simple Concerned.”  I suppose “Sleepless in Seattle” had already been used.
     
    Vinny

    0
    #118858

    Nwajei
    Participant

    You still haven’t answered his questions!!  What is a better tool.

    0
    #118886

    Six Sigma Tom
    Member

    Gunter also wrote a column in Quality Progress entitled “Farewell Fusilade” where he proposed that the statistical methods used in the quality profession were out of date. Same argument I’m making today for the same set of tools.

    0
    #118889

    Mikel
    Member

    You are not making an arguement, you are talking in vauge generalities. Tell us specifics.
    I will make you a bet that I can take any problem and use the tools that you call old fashioned and flawed and solve the problem quicker and better than you and any 10 new age stats guys can do. You are just a bunch of theoretical fluff.

    0
Viewing 75 posts - 1 through 75 (of 75 total)

The forum ‘General’ is closed to new topics and replies.