MONDAY, MAY 28, 2018
Font Size
Topic Gage RR: How many samples, operators and repetitions?

Gage RR: How many samples, operators and repetitions?

Home Forums General Forums Tools & Templates Gage RR: How many samples, operators and repetitions?

This topic contains 81 replies, has 13 voices, and was last updated by  Mike Carnell 11 months, 2 weeks ago.

Viewing 82 posts - 1 through 82 (of 82 total)
  • Author
  • #167690 Reply

    Hey all,

    I was thinking about this today, and realized I know less about it than I thought.

    I know the rule of thumb answer to this question is to take 10 samples of similar spread as what you are going to be measuring in the process, using three operators and two repetitions, but this just isn’t satisfactory to me.

    Can anyone explain where this rule of thumb came from, and speak to a methodology for calculating my own sample sizes? Is it similar to a power analysis for a normal distribution?

    I am particularly interested in this as the FDA regulations require us to provide a statistical rationale, and I am at a little bit of a loss here.

    Thank you for your help,

    #167795 Reply

    The two main types I have worked with are the Type 1 1x1x50 and the Full Gage 3x3x10

    [The remainder of this post was removed due to bad info]

    #167797 Reply

    This “rule of thumb” comes from AIAG guidelines – you can inform FDA not to expect any rationale from them ( i.e The AIAG) ! For ex: Can anybody explain the rationale behind NDC ?

    The Gage R & R analysis is a messy affiar ! It has been ‘designed to confuse’ not ‘designed to conform’ . High time we simplified this method. The statistical software makers are quietly cashing in on the prevailing chaos

    #167808 Reply

    Wow bbusa,

    Your answers aren’t usually this ignorant.

    I suggest you read the Minitab documentation and Wheeler’s Measurement System Evaluation.

    Stop expecting to be spoon-fed.

    Hint – NDC and the stdev of the measurement system are related. See if you can figure it out from there. NDC and Wheeler’s Discrimination Ratio are also basically the same number

    #167812 Reply


    Thanks for the back handed pat !

    You have’t answered the original post .

    I know how NDC & SD are related . I also know SDs should not be added . Why do we need so many different metrics when a simple % variance will do the job ?

    About spoon feeding : is it for me or the orignal poster ? If it is for me , it is the last thing i expect from you !

    #167814 Reply

    Stan’s views on Wheeler : posted 1 year 8 months ago on this forum


    Wheeler is a buffoon – his advice has no value in the real world
    where we use these measurements to accept or reject product.He would argue that we use evidence of control and capability
    which is correct but not reality.The Gage R&R should not be part of the study of variation, the
    contribution of the measurement should be known to be
    acceptable when introduced into a facility – it should never be a
    question at the time of usage.Andy U’s article is quite good and useful and shows the value of
    Shainin’s Isoplot as a way to visualize measurement error. It doesn’t
    address Wheeler’s points because they are red herrings.Wheeler has been beating this same point for two decades and has
    only succeeded in making control charts less sensitive and
    processes less capable.

    Unquote :woohoo:

    What Don Wheeler had said in his article ” Honest Gage R & R Study”


    ” ……After Step 6, every aspect of the Gauge R&R Study is erroneous, fallacious, naive, incorrect, and, to put it quite simply, wrong .
    ………. Clearly, the main purpose of a Gauge R&R Study is to condemn the measurement process,and to achieve this end the summary measures are inflated so as to overstate the damage of measurement error while the guidelines used are excessively conservative. So what can you learn from the AIAG Gauge R&R Study? Virtually nothing that is true, correct, or useful! You have taken the time and gone to the trouble to collect good data, and then you have wasted the information contained in those data by performing a hopelessly flawed analysis.
    So what should you do? The least you can do is to stop using the AIAG Gauge R&R Study and start using an Honest Gauge R&R Study, as described in the next section


    #167818 Reply

    Again bbusa, let’s use your brain this time.

    That I suggest you read Wheeler doesn’t mean I agree with his interpretation of rules surrounding GR&R.

    Wheeler’s take is from the point of view of everything is done to study a process, but those of us in the real world need to worry about product acceptance as well.

    Read Wheeler for a very good development of Discrimination Ratio which for all practical purposes is NDC.

    I believe you read broadly to learn, not read broadly to be a tool zombie. You have to think, you have to put things into the context of your own experience.

    I like Wheeler a lot but my experience says to follow the rules of AIAG (and yes figure out why). I have seen about 25% of projects over 30 + years of experience be clarified and given direction by getting acceptable results per the AIAG rules. If I had followed Wheeler’s advice I’d still be pontificating my navel on the first one.

    Wheeler’s advice on transforming data is pristine by the way. Breyfogle is the buffon on that front.

    #167819 Reply

    Stan :

    I suggest you not to live off other ‘ s brains , instead check if your’s is in the right place .

    The amount of bull#!$# you post on this site – is second to none . I am not sure if you are being paid to post .

    First you called Wheeler a buffoon , now you call Breyfogle a buffoon , what are you – surely worse . In my opinion they have acheived & contributed far more than you can even think of. 30+ years of exp – i don;t believe it !

    Take my advice : forget my brains , try and answer the original post .
    Atleast it will help someone in need . Before doing that read Wheeler’s Honest Gage R & R again .

    #167820 Reply

    Hey bbusa,

    Read your answer to the original question. What ignorant BS.

    I read all that Wheeler publishes and he is wrong about Gage R&R.

    I don’t post BS I call out those who pretend to know who don’t. I also answer questions where it is clear the poster has done their homework first. I gave the original poster information where they can become more informed than most (like you) if they choose to do so.

    Your answer on Destructive R&R was dead wrong. Why don’t you come back with something clever on that too?

    #167828 Reply

    Matt: While the kids squabble, I’ll try to help you out. I can’t speak to how the rule of thumb numbers were established, but just logically you need some spread of sample values that represent the spread of the measurements to be taken, 10 seems a reasonable number to cover the spread (2 low, 2 high, 2 mid-range low, 2 mid-range high and 2 around the middle). 3 Op’s is a minimum and should only be used for an initial MSA. Should your measurement system not be very highly capable then you should perform the MSA with all the operators. In the event that this becomes overwhelming, then observation and selection of those least capable (check the measurements being taken by operator and identify those deviating the most from the majority, either high, low or range). The intent of the MSA is to ensure that your measurement system is not unduly clouding your observed data. You will want to ensure this is the case for the least capable operators. Your statistical rationale in this case would be that these operators showed the most variation in readings (how would you evaluate a variable like operator with multiple levels – people – having a continuous response variable – the measurement?)
    Hope this helps.

    #167829 Reply


    You read all that Wheeler publishes and then call him a buffoon . You must be mentally sick . You can t be asking me read Wheeler and at the same time saying that he is wrong about Gage R & R

    Isn’t it like asking you “Have you stopped reading porn”?

    Anyways , I am still expecting your response to the original post : the statistical rationale behind GR & R sampling

    #167830 Reply

    I read all sorts of things from people I don’t agree with. I only disagree with Wheeler on his interpretation of results. And if you read what I wrote you already know why.

    I gave the original poster a reading assignment which will answer his questions. It’s his choice to learn.

    You must be stupid.

    #167832 Reply

    Here ‘ s another article that talks of inaccuracies in Gage R & R study .
    Unless , of course , Stan calls this gentleman a buffoon !

    P.S. : Unable to attach the article as the size exceeds 120kB . The article is ” Improved Gage R & R Measurement Studies by Donald Ermer , Source : Quality Progress March 2006 .

    Excerpts :

    Change Is Needed :
    AIAG R&R methods may be misleading
    and should be modified.
    Reliable measurement data and analysis
    of those data are important. Part
    one of this column will help conscientious
    organizations further improve
    the quality of their products and the
    productivity of their processes.
    In part two, I will introduce the
    appraiser variation (AV) as the third
    component of the total product measurement
    variation and will give an
    example of a complete and correct
    gage R&R measurement study.

    I am sure many practioneers will appreciate the need to simplify the Gage R R method to ensure that the shopfloor supervisors/ foremen are able to run the process themselves without relying on self – proclaimed brilliant consultants !

    #167834 Reply

    From your original response –

    “you can inform FDA not to expect any rationale from them ( i.e The AIAG) ! For ex: Can anybody explain the rationale behind NDC ?

    The Gage R & R analysis is a messy affiar ! It has been ‘designed to confuse’ not ‘designed to conform’ . “

    Rational behind the NDC – read Wheeler’s explanation of Discrimination Ratio. The only difference is a -1 under the square root. This was done to simplify the calculation. The difference in practice in noise.

    Gage R&R was designed in 1962 by a group of engineers at GM. The choice of sample size and also of confidence interval was a compromise between statistical validity and what could be implemented to a fairly statistically ignorant engineering population. The rules based on standard deviation instead of variance are an artifact of using Range as a way to minimize computation error when this was done with pencil and paper (1962).

    Should the Range method be abandoned? – yes and it has for the most part.

    Should we use variance instead of standard deviation? – it doesn’t really make any difference if you understand how we come up with the rules. For simplicity and statistical understanding, variance is more straight forward. But the rules based on standard deviation are just the square root of the rules based on variance.

    And on consultants – wouldn’t it be nice if they at least understood what they teach? I laugh when I see people teach that the rules on NDC, variance and standard deviation all must be met – they are the same rule.

    bbusa, I have a quiz for you.

    What is the right rule for %Study? 30% or 9%. Pick one and tell us why.

    I promise to stop if you can answer that correctly.

    Any idiot can do a web search and post links (or in your case, not post links). >50% of what gets published in Quality Progress is tripe and has been for at least 30 years. Tell us the point of the article and what you learned from it. Oh, you haven’t actually read the article?


    #167843 Reply

    “The choice of sample size and also of confidence interval was a compromise between statistical validity and what could be implemented to a fairly statistically ignorant engineering population.”

    What a confession ! You seem to be coming around .
    That’s precisely why I said there is no statistical rationale .

    “For simplicity and statistical understanding, variance is more straight forward. But the rules based on standard deviation are just the square root of the rules based on variance.”

    I think 3(squared) / 5 (squared) is not equal to 3 / 5 !
    So mathematically , they are not quite the same

    I laugh when I see people teach that the rules on NDC, variance and standard deviation all must be met – they are the same rule.

    So shall we request the Statistical Software makers to redesign the tool ? Else , how do we convince ourselves that a threshold of 9% on variances is equal to NDC of 4 ( sorry now revised to 5 )

    No, I don’t want you to quit posting . You bring in the much needed oomph factor to this forum ! BTW , what ‘s % Study ? I thought the study of variances was more straightforward !

    #167844 Reply


    Man are you stupid.

    No confession, a compromise is not the same as no rationale. For example, sample size selection is always a compromise.

    The rules for % contribution are the square of the rules for %study. You might want to at least do the hard work of using a calculator for ten seconds before posting stupidity that took longer.

    The statistical software makers are only reflecting what is written by AIAG. And yes, everything written by AIAG is a committee decision.

    What is the answer to my question? You don’t know, do you?

    #167846 Reply


    Aren’t you a super idiot ?

    Basic math tells us SDs cannot be added . That’s the reason % variances is more logical ( < 9%)

    You may have any no of rules , but that should make sense . Your question on % study is a no brainer !

    #167847 Reply

    What is the answer?

    #167852 Reply

    Define % study

    #167853 Reply

    P/TV – the ratio of stdev of the ms to stdev total

    It’s called %Study in Minitab these days

    What’s the answer to the no brainer question?

    #167854 Reply

    < 10 % excellent , < 30% acceptable

    Still a no brainer – as you are comparing SDs rather than variances

    #167868 Reply

    There is no difference. Why are you hung up on it?

    #167869 Reply

    The MSA is a study of components of variation . What sense does it make if all of them ( the various components ) don’t add up to 100% ?

    #167870 Reply

    But they do. Variances add, standard deviations don’t. Square the results of %Study and guess what they add up to?

    Guess what NDC and Discrimination Ratio are the ratio of – standard deviations or variances?

    Why is 30% acceptable and how was 30% derived?

    #167871 Reply

    BTW, if you were to use the results of a study to guard-band your specs, which would you use? Variance or Standard Deviation?

    #167872 Reply

    lighten up stan.

    i heard it was <20%. why? because it is a little better than <30% but not quite as good as <10%.

    now please move on, you are clearly the less stable one in this fight!


    #167873 Reply

    Mr. t,

    You don’t have a clue.

    I am trying to help bbusa learn something. Maybe you should THINK instead of come back with an empty remark.

    #167874 Reply

    Mr. t., I like it. Thanks for the respect.

    You helping bbusa? Give us a break. Based on your tiresome and useless rambling babble it seems to me you are not capable of helping anyone. Just a guess, but I am thinking you spend a good deal of your time alone.

    btw, my empty remark is a true statement.


    #167875 Reply


    Thank you for helping me to learn – but i want to learn the right things !

    I would like to know from you , if the study of variances , % study , NDC are all the same , why have we created so many different metrics – to confuse the engineer ?

    This is the precisely the issue I raised in my first post . Surely , there should be some advantage / difference of one metric over the others

    Also , try explaining the concept of NDC to the shopfloor supervisor , who is responsible for Gage R & R studies !

    I still do not understand your claim that the %study adds upto 100% . Can you show me any Gage R & R result ( look at the example provided in Minitab) to prove this

    #167876 Reply

    The four metrics –

    P/T also known as %Tolerance – the original and still probably most important. It used standard deviation derived from range. I’ve already told you why range was used when this was created in 1962.

    P/TV also know as %Study – the one of interest to those trying to improve processes. You need to know if your measurement will differentiate improvements and not be an important source of variation. It was created prior to all the software available these days and used standard deviation derived from range.

    %Contribution – created for the anal statisticians after software and the ANOVA method (which is better information BTW) came into vogue. They, as you, are hung up on the numbers not adding. Every number reported is simply the square of the number reported for %Study. So that’s my comment about %Study – square each number and add them – 100%!

    NDC or Discrimination Ratio – The Wheeler camp. Uses the ratio of standard deviation of ms divided by standard deviation of the rest.

    Why all 4? – The first two are for the people who learned this years ago and have used it effectively. The third for for the anal stats guys and the fourth is for the anal Wheeler camp. AIAG is a consensus spec and they try to accommodate everyone’s opinion. They are the standard in this case.

    Mr. t is a real idiot isn’t he?

    #167877 Reply

    BTW – I only teach right. I don’t teach anything I don’t understand – like the idiotic comment about 20% from Mr. t. I know why 30% was put out there 48 years ago and I know the idea behind it is correct.

    #167878 Reply

    Stan, thanks for the shout out.

    Lets see what our data tells us about you so far:

    1. Stan is not cable of teaching anyone anything.
    2. Stan does not listen well.
    3. Stan appears to ignore good advice.
    4. TBD

    What will our next data point reveal. That Stan is a relic that has lost sight of the benefit of learning from those that are more well rounded than himself.

    While we wait for that next data point I will continue to ponder what makes you tick. Do they keep you locked in a room. Are there bars on the windows, padding on the walls, and people in white jackets guarding your door 24/7.

    Life is full of surprises,


    #167879 Reply

    Mr. t,

    Any idiot can just take shots at people and you are proof of that.

    Do you actually know anything?

    If you do, give use a data point to your four idiotic points –

    1) I’ll bet I have and continue to teach more people than anyone on here.
    2) I don’t listen to idiots, but that is only a small lunatic fringe which you seem to be a member of. Seems as though I have been listening to bbusa and answering his questions in a socratic fashion.
    3) What good advice – give me even one good example.
    4) Was there a point here?

    This thread is about MSA – tell us what you know and how you have used it. Or have you already exhausted what you know?

    #167880 Reply

    Holy S*** !!!

    What do we have here – A monster

    1. One for the old foggies
    2. One for the improving processes – as if improvement is not possible with other ratios
    3. One ( the most statistically & logically straight forward) for idiots like me
    4. One for Wheeler

    This is ludicrous , to say the least . AIAG has never in the past 3 decades explained the logic behind their naivity in such simple terms . I hope you do not represent them . No wonder our cars cost so much more !

    Stan , I thank you for the enlightenment .

    I urge all practioneers not to confuse themselves anymore and use one simple metric i.e %contribution . There is enough information in this metric – to improve the MSA

    #167881 Reply

    I almost agree.

    % Tolerance or P/T is still important. It is possible to have a measurement adequate to improve a process that does not protect the customer. This can happen any time Cp<1.

    For process improvement purposes %Study or %Contribution are both okay since they are just manipulations of the same numbers. NDC or DR are particularly insensitive since you round down to zero significant digits.

    #167882 Reply

    BTW – no monster.

    1) The old guys gave you the opportunity to do what you are doing. We fought to get MSA accepted. We worked with the software makers to assure agreement with the most common standard – AIAG. We made sure %contribution was included because we as well as the guys who created this in 1962 understood the underlying principles. We wanted to make sure the masses who were being trained would understand as well.

    2) Improvement is certainly possible with other ratios but we wanted to transition the folks who were successfully using P/T. That’s an odd thought don’t you think? Work with the folks who are making use already instead of the snot nosed kids who aren’t doing anything yet.

    3) You are only an idiot for making a big deal out of something that has been understood for 48 or more years. Software makes a better understanding possible to a wider audience, it doesn’t make anything statistically or logically more straight forward. I know the guy who first did this at GM and he could have explained all this to you in 1962. They made logical decisions for the time – by using range, they made this accessible. The 30% rule is also a “compromise”. It should be 31.62% – do you know why and why they chose to make it 30%? Also the 9% rule for % contribution is wrong – the people who came up with that did not understand and fixed the wrong problem.

    4) Wheeler’s book on measurement systems is good, but he only addresses the needs of studying a process, not the need of using measurement to protect the customer. Folks who follow his advice fail to improve measurements, they just learn to study their processes in spite of their measurements.

    If you only use %contribution, how will you know if your customer is protected?

    My opinion – you need to be able to know the status of the gauge regardless of the language being spoken. I can tell you % contribution if told % study. I can tell you % study if told % contribution. I can tell you the range of %study or % contribution if told NDC or DR. Why wouldn’t that be your objective? To do this means you really understand and have gotten away from the crybaby %contribution rant you seem to like.

    #167883 Reply

    Stan : in your last post you’ve come out as a bundle of contradictions . As far as 31.6 % , 30% , 9 % confusion is concerned , you can shove these numbers “you know where” .

    If you only use %contribution, how will you know if your customer is protected?

    If your % contribution shows 20% and you are still shipping products , you need to protect your backside first ! Don;t worry about other numbers

    You have begun showing symptoms of a rare disease that borders on extended childhood & premature senility . You have yet to decide if Wheeler is a buffoon or someone you would like to follow .

    Stan : Lead , Follow or get out of the Way ! Choose one . Your opinion is nothing short of stupidity .

    It is high time we tamed this monster called Gage R & R that you and your likes have created – in cahoots with the software makers . That will do a world of good to the industry

    #167884 Reply

    Folks, we have our 4th data point. Thanks Stan.

    Data point 4 is a quote from one of Stan’s most recent posts;

    “1) I’ll bet I have and continue to teach more people than anyone on here.”

    Lets review our CI Save Stan Project Plan,

    Define – Our Project Charter.

    Problem Statement – Stan’s behavior consistently derails his message preventing him from sharing his wealth of knowledge in an effective manner.

    Project Description (Goal) – Develop a strategy that Stan can utilize to fulfill his desire to interact with others that will ultimately improve the abilities of his audience.

    Scope – Anyone that has behavioral traits similar to Stan.

    Measure – Identified data points that reflect Stan’s behavior.

    1. Stan is not cable of teaching anyone anything.
    2. Stan does not listen well.
    3. Stan appears to ignore good advice.
    4. Stan believes he is all knowing and cannot learn anything from anyone.

    Analyze – Review our data and identify options to ensure that Stan’s bus is always filled with passengers that are excited about seeing the same sights and reaching the same destination as Stan. Our options are as follows:

    1. Ignore Stan’s behavior.
    2. Accept Stan’s behavior.
    3. Determine the root cause of Stan’s behavior and provide him with simple corrective actions to improve his behavior.

    Improve – Options 1 and 2 are certainly the easiest to implement and very appealing to some, but we concluded they would not benefit anyone. Stan would have to continue driving around in an empty bus which is just plain unacceptable.

    That being said, we unanimously selected option 3, because ultimately we really do value Stan’s opinion. Our root cause analysis revealed that by focusing on correcting data point 4 first, data points 1, 2, and 3 will naturally and automatically resolve themselves. Our suggested data point 4 corrective action plan is as follows:

    1. Convince yourself that learning from others helps everyone including the bus driver.
    2. Move on to another post. This bus is broken down causing traffic to come to a stand still.
    3. Initially, do not be the first one to respond to new postings. Give the other bus drivers a chance to pick up new passengers once in awhile.
    4. Do continue to read all postings and replies. Stay focused on posts that have a reply where you can add value to both poster and responder. Resist responding to posts where you have nothing valuable to add.
    5. Start all of your posts by complementing the responder. Nothing complicated, just a simple “Nice point bbusa or I never looked at that way.” Then begin delivering your message “Have you ever looked at it from this angle or I have seen that method work, but with a slight twist.”
    6. This step is your “Herbie” and is the most critical step of your corrective action process. After you have typed your response review it carefully and remove or reword any statements or connotations that have even a hint of a tone similar to the following statements:

    a. Your answers aren’t usually this ignorant.
    b. Stop expecting to be spoon-fed.
    c. let’s use your brain this time.
    d. What ignorant BS.
    e. You must be stupid.
    f. Any idiot can do a web search and post links.
    g. Man are you stupid.
    h. You don’t have a clue.
    i. like the idiotic comment about 20%
    j. Do you actually know anything?
    k. snot nosed kids who aren’t doing anything yet.
    l. the crybaby %contribution rant you seem to like.

    7. Always end your replies on a positive note. Again, nothing complicated, just a simple “This can be a tough topic for some but you are right on target” will do.

    Control – As I mentioned earlier, we really do value your opinion and are confident your bus will be full of excited and well educated passengers in record time. So going forward, if you do not hear from me, this will be a good thing, as we will all know that Stan’s Bus is in control.


    #167885 Reply

    Wow Mr t, what lovely advice. I forget sometimes that the only thing important in life and by extension Six Sigma is that we all hold hands and skip together.

    What BS.

    Do you actually know anything about the topic of Gage R&R?

    No? You probably ought to move on to the kindergarden forum.

    One thing I do know from your post – I thank God I don’t work for a weak kneed person like you. I’ve always fired bosses with similar approaches.

    #167886 Reply


    You don’t actually understand the topic, do you?

    If you have % Contribution of 20% but Cp, Cpk, and Cpm > 2. Who care about the 20%.

    If you have %contribution of 1% but Cp=0.1, your customer is at risk and your measurement system is inadequate.

    Do the math, you don’t know what you are talking about.

    #167887 Reply


    You’ve lost it completely .

    If you have used the same measurement system ( of 20% ) to declare Cpk of 2 , you would be called a – –

    #167890 Reply


    You need to review the stats behind %contribution. It is absolutely possible to have %contribution of 20 with a Cpk of 2, 4, or even 20.

    Do your homework before you respond. You don’t even understand the number you claim to be passionate about.

    #167891 Reply

    I have done my homework as always – you need to do yours and not throw numbers around

    Everyone here now knows how much you understand

    #167892 Reply


    Who knew people got into heated debates over Statistics…

    I guess you can have an argument about pretty much anything on the internet.

    I’m surprised noone has invoked Godwin’s law yet.

    #167893 Reply

    He guys please don’t argue and concentrate on the topic.

    #167895 Reply


    You have not done your homework.

    You don’t understand the relationship between %contribution and Cp/Cpk.

    Hint –

    standard deviation total = square root of (standard deviation of part/process squared + standard deviation of measurement system squared)

    % contribution = standard deviation of measurement system squared / standard deviation total squared

    Cp = Specification width / (6 * standard deviation total)

    Note that the standard deviation total is the same number in all equations. Play with varying the standard deviation total and holding spec width and standard deviation of measurement system constant. A good %contribution can be an artifact of an awful Cp. A bad % contribution can be an artifact of a very good Cp. Cp and % contribution (or % study) have to be considered together to know if the MS needs work.

    #167896 Reply

    I am concentrated on the topic.

    Mr. t is just plain dumb and bbusa doesn’t understand.

    #167898 Reply


    Hint ! This is schoolboy stuff . You think you sound brilliant posting formulae , which evena beginner would know

    All I said is if your gage r & r is 20% , you cannot claim your Cpk is 2 , as the measurements would have been taken using the erroneous gage or operators or both !

    It is you who refuse to understand – by your own admission , you should have quit this forum after I answered your NO BRAINER .

    You are at it again – posting silly senile stuff over and over again .

    #167913 Reply


    If this is schoolboy stuff, you must not have done well in school.

    %contribution is % of the total variation. 20% of total variation if your Cp is 1, 20% if your Cp is 2, 20% if your Cp is 200.

    Maybe a beginner would know, but you don’t understand at all.

    All of your noise about preferring the statistically correct %contribution should be spent on understanding what it means. You are just parroting what some other buffoon taught you. They did not understand, you do not understand. Maybe you should teach someone to keep the stupidity going.

    #167915 Reply


    I have very well in school FYI – maybe better than you .

    As usual you are talking rubbish . All you know is to call others buffoons , stupid and what not . You are trying to bully others with your senile stuff , it is not going to work

    I may be quoting others , but I don;t call them names . It is high time you corrected your behavior

    You are simply fooling around with the % contribution , % study stuff – that doesn’t make you look a bit intelligent . Tell us something different ,else get out of here as you promised – do you keep your word ?

    #167925 Reply


    I’m not going anywhere.

    It’s time for you to stop talking and start thinking. You are just flat wrong.

    Take the time to go understand the relationship between %contribution/%study and the capability indices.

    You can have an acceptable %contribution with a bad capability. If you improve the capability, your % contribution number will be worse and at some point be unacceptable. When you understand that, you’ll understand the rest of what I’ve been trying to tell you.

    #167971 Reply

    Reputation - 0
    Rank - Aluminum

    bbusa and Stan,

    Interesting the passion around this topic. Although I do not agree with Stan, his understanding of the topic is correct.

    bbusa, please look at my derivation of Discrimination Ratio and understand how the numbers work in the computation of Capability.

    #167975 Reply

    Reputation - 0
    Rank - Aluminum


    When we talk about Gage R&R we are talking about Repeatability & Reproducibility So why do why have multiple parts and multiple test. Simple terms this is why.

    Repeatability is the variation between successive measurements of the same part under the same conditions. It is used as an estimate of Short Term Variation. In order to accurately capture this, one must measure the sample more than once, 3 is considered the minimum for best results.

    Reproducibility is the difference in the averages of the same samples using different people or even different gages.

    Accuracy and Precision are often confused as well, Accuracy is the measurement system bias as related to the actual or true reading. Precision is in reference to measurement system variability and standard deviation of the measurements.

    Some rules of thumb:
    1) Samples should be representitive to the process (not hand picked)
    2) Minimum of 10 samples (Some exceptions to this rule, ie expensive or destructive test)
    3) 2 operators or 2 different gages.
    4) Look at your # of distinct categories, Not having enough samples or number of measurements can lead to low number, Should have at least 5 distinct categories in your samples for process improvements. for any study you system should be able to divide samples into at least 3 distinct categories.
    5) Should always repeat measurements in the same order (Standard Deviation error can occur if you do not do this) As a personal rule of thumb, I always measure samples 3 times with at least 2 people and often use 3 different people.

    For some more information go to or check out Moresteam under MSA

    The answer to your questions is right in front of you, what are trying to do with an R&R? Repeat & Reproduce, how do you do that? Repeat the measurement of the samples, then see if you can reproduce it with different people.

    Chad Vader

    #167980 Reply


    Thank you . I have read your derivation of DC .

    The point I am trying to make is – why do we have so many ratios & %ages when each one is either derived or approximated using the other . The basic measure is the % variance – that should be enough .

    Surely , there should be some logic to this .

    #167982 Reply

    Wow – not sure if you’re “BBUSA” as your letters are switched around in your nom de plume (I’m guessing not, so you’re probably one of the perpetual thorns on this board).

    Anyway, the situation you and Stan are arguing about is one of the “trick” questions that I always give my BB candidates and interviewees. If you really knew what you were talking about, you’d have a different answer than the one you keep espousing here. Try again (or maybe get your money back from whomever sold you your fake certificate).

    #167983 Reply

    Stan: I think you’ve been taken to the rodeo and rode hard by a BBUSA impersonator.
    Plus, how in the world have you maintained a Karma of 0?

    #167999 Reply

    The only thing I can figure is karma gets reset to zero when you hit -6.

    Either than or bbindia (bbusa who claims to be from India on his profile) really likes me.

    #168000 Reply

    % variance changes every time capability changes – how can it be the one measure when it’s always changing?

    #168009 Reply

    If % variance variance changes “everytime” , then % study also changes , same logic applies to all other BS metrics – NDC etc

    #168010 Reply


    You may want to stop, think, and reread this thread.

    You are the only one who is advocating a single, end all, be all metric.

    You presented the thought that if you achieve a %contribution of x% (you fill in the number you think is appropriate) that it is all you need to know.

    All I’ve tried to point out is two things –

    – %contribution (or %study or %process or NDC or DR) are points in time and they only tell you if your present measurement capability is capable of helping you improve your present process capability. Improve your process capability and that same measurement system may be incapable of helping you improve further.

    – Your measurement system’s ability to discriminate good from bad is also important in many cases. In this case %tolerance with it’s legacy rules of < 10% and < 30% need to be understood in conjunction with process capability and guard-banding. I explained to you how we got the legacy rules and why it doesn't really matter if you understand them.

    You started this thread with thrashing the thinking that has led to the point of having 5 metrics to tell you about your ability to study a process that all tell you the same thing. I agree with you that it’s dumb but it also doesn’t matter if you understand the relationships. What else that is dumb is the person that has a rule for several of the 5 metrics and the rules don’t line up.

    You have failed to acknowledge that a measure’s ability to discriminate good from bad is also important. In this case there has only been one metric since 1962 and the method for computing it and breaking it down has only evolved once – from the range method to ANOVA. This occurred when computers and appropriate software became ubiquitous – I personally know it happened with Lotus 123 in the late 80’s, but maybe is happened with Visicalc a little earlier.

    The practice of declaring a study with a %contribution of x% adequate forever is irresponsible. If you want to do such a thing you need to use desired variation in the denominator instead of existing variation – that would work, but I don’t think that is what you are advocating.

    #168011 Reply

    The only ratio that stands up is gauge squared / process squared. All the others are driven by that one. So who cares if they add up to 100% as long as you understand how to interpret the %’s? Incidentally, you can calculate any index from any other except ndc because it’s truncated.

    As for the argument about capability, tolerance and capability index drive required proces variation which drives required measurement system variation which drives maximum resolution.

    #168020 Reply


    This is clearly your most logical post on this thread – I might disagree on some points , though

    If you are looking at improving process capability , you should first improve the measurement system , not the other way round ( in the least ” buy” a better measurement system . Without a capable MS , there is no way you will establish process capabilities

    By % Tolerance : I assume you are referring to P/T – If yes , I don’t think it is an intelligent measure – for the simple fact that it is prone to varying customer specifications . In many cases the tolerances set by customers are questionable – atleast that’s my experience from the numerous DFSS /NPD workshops that I have conducted . Tolerancing is a big, big problem / issue

    Yes, a better approach is to study the corresponding Xbar R chart along with the % contribution results .

    If you think it is dumb to set rules for all 5 merics , just check what the TS auditors are doing – how many really know what NDC is and how it is related to % contribution ?

    #168021 Reply

    It seems to me that people are not seeing the wood for the trees and not answering the question.

    Nobody asked what Matt wants to do with the measurement system. And I don’t see how to answer the 9% / 30% question without knowing what the gauge will be used for. Will you be controlling a process or improving it / correlating with another gauge?

    How do I know how many samples, operators and repeats to put in my MSA? Follow convention and go from there. 3 – 10 – 3 will be enough to show you if your measurement system will characterise your process as represented by the parts. Always include tolerance in your analysis if it’s possible. That way you will avoid improving a gauge you don’t need to or not improving a gauge you do need to.

    I hate to come between a man and his wife but all this barren academic mud-slinging will bring us into disrepute. It’s almost as bad as Yahoo! comments.

    #168023 Reply

    I can see the conversation –

    Customer – We’ve measured your parts and they do not meet the specifications your company agreed to. Do you know your ability to measure these parts?

    bbusa – I assume you are referring to P/T. I don’t think it is an intelligent measure.

    Customer (to bbusa’s boss) – Who is this fool?

    bbusa – I don’t pay any attention to specification, My experience from conducting DFSS/NPD workshops is the tolerances you set are questionable.

    Customer (to bbusa’s boss) – Who is this fool?

    bbusa – Tolerancing is a big, big problem.

    Customer (to bbusa’s boss) – Who is this fool?

    bbusa’s boss – go pack you things bbusa. I am sorry Ms. Customer, we thought this guy was marginal but could be salvaged. He told us our measurement system was good but it you will give me a half hour I’ll set up a study to see our ability to judge the specifications we agreed to with you. We’ve been having a lot of problems since we started hiring these people called BB’s, they seem to have a very poor understanding of basics.

    #168024 Reply

    By the way , this fool is none other than Stan

    What else would you call a fellow who claims to have 30 years of exp but thinks of 5 different metircs to explain the same measure – just in case one goes wrong , he can show the other to the customer and get away – The Big 3 Style !

    A dog’s tail can never be straightened – I was wondering what went wrong with your post yesterday – maybe the dog had its tail between its legs !

    #168025 Reply


    Your post about tolerances and %Tolerance (yes P/T) was close to being the dumbest post ever on this forum – and that is a real low bar.

    If you don’t believe in specs, you don’t believe in capability. If you don’t believe in capability, there is no need for MSA or % contribution.

    I never said I needed 5 metrics, I just told you I understood all five and don’t get anal about any of them. That’s four less than you get anal about. No sorry five since you don’t believe in tolerances.

    #168027 Reply


    To enlighten…pl read Montogomery’s SPC pages 379 – 380 – why the P/T is not a great measure

    i.e if you wish to update yourself and stop beating the 1962 GR & R BS

    #168030 Reply


    Your charade has gotten more than a little old.

    If Montgomery doesn’t think knowing you measurement capability with respect to you accept / reject criteria, he clearly doesn’t have a clue. The same goes for you.

    #168031 Reply


    It will be worth getting off your high horse and going back to the basics . Your brains badly need some 5S

    #168034 Reply


    Knowing your measurement capability vs your accept / reject criteria is being on a high horse?

    Dude, this charade has gone on too long…

    #168035 Reply

    Who is this Darth you are talking to ?

    …what is this accept / reject criteria you are referring to ? I thought your brilliant NDC / % study will take care of that

    #168037 Reply

    Reputation - 0
    Rank - Aluminum

    If your offended by this post then its probably meant for you……

    I am amazed by people that read a book, have just enough experience
    to be dangerous, and then begin to develop their own theories without ANY
    practical knowledge or understanding……

    Where is Stevo when you need him

    Chad Vader

    By the Way BBSUA, the customers tolerance is never a Big Problem nor is it a small problem. Competence to have the correct measurement system in place for the customer specification is the problem. Competence comes with knowing how to use tools such as P/T, and applying the correct evaluation of process capability along with it.

    #168039 Reply

    bbusa, bbindia, Darth, or whoever you really are –

    OK, I give up. You really are that stupid.

    #168042 Reply

    Chad Vader

    Thanks for the verbose . Nothing on this forum is offensive anymore – thanks to people like Stan

    I don’t understand what you meant by :

    “the customers tolerance is never a Big Problem nor is it a small problem.”

    Quite funny . Is it from some book again ?

    The rest of your ‘prose’ is as bookish as it could get

    #168043 Reply


    It is your brain dead approach about 5 different metrics for MSA that can be called stupid .

    #168053 Reply

    Reputation - 0
    Rank - Aluminum

    Honey, (bbsua)
    Did your mom drop you on your head? Bless your heart.

    My comment about customer tolerance came from your post stating that customer tolerance(s) were a “Big Problem”, not my words, yours amigo. Your implication that somehow the customer is to blame for poor quality because of their tolerance is just plain incompetence. My statement was merely another way stating that the customer is always right, deal with it.

    #168054 Reply


    We are discussing MSA here

    Customer is always right is a very general statement .
    Stop sleeping with Stan – it is affecting you IQ

    #168056 Reply

    Reputation - 0
    Rank - Aluminum

    What I’ve seen here is Stan, along with others discussing MSA and you spouting off at the mouth like a 15 year old know it all kid, using mis-interpreted information to push an agenda of what seems to be turning this site into another Yahoo play ground.

    Judging from your comments, your just another academia trying promote something of which you have no real world experience about. I am sure you have been on the site before as your “prose” wreaks of past idiots before you.

    Good luck with your customers, I’m sure they are very happy with you. You now have this thread all to yourself big boy.

    #168057 Reply


    Talk about MSA , else come back after you have read about it

    #621700 Reply

    Hi Guys,

    May i know the confidence level of using the 10 sample size in Gage R&R?and how they derived it?

    Hope your soonest respond on my inquiry.


    #621983 Reply

    Your confidence level also depends on the number of “operators” and repeat measurements. Follow the AIAG guidelines and you can’t go wrong.

    #627695 Reply

    @andrada I can’t believe you dragged that up from 7 years ago. Those were the “good old days” of what Bill Hathaway at Moresteam refers to as food fights. There was more than a little deja vu (not the movie) reading through all of those old posts.

    Just in case you are wondering I have worked with Stan/Mikel for several decades and he is pretty dead on in terms of what he knows and teaches as is @cseider in his link.

    It is a less caustic world these days but a lot less fun and a lot less informative. People may not like that exchange but there were a lot of good issues put on the table.

    Just my opinion.

Viewing 82 posts - 1 through 82 (of 82 total)

Register Now

  • Stop this in-your-face notice
  • Reserve your username
  • Follow people you like, learn from
  • Extend your profile
  • Gain reputation for your contributions
  • No annoying captchas across site
And much more! C'mon, register now.

Reply To: Gage RR: How many samples, operators and repetitions?
Your information:

5S and Lean eBooks

Six Sigma Online Certification: White, Yellow, Green and Black Belt

Six Sigma Statistical and Graphical Analysis with SigmaXL
Six Sigma Online Certification: White, Yellow, Green and Black Belt
Lean and Six Sigma Project Examples
GAGEpack for Quality Assurance
Find the Perfect Six Sigma Job

Login Form