2 Gage study MINITAB question

Six Sigma – iSixSigma Forums Old Forums General 2 Gage study MINITAB question

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
  • #38177


    I am comparing two gages. I have 1 operator and 10 parts. Each part was measured twice with each gage. How can I compare the data? Previously I ran another study with only 1 gage, 2 operators and 10 parts. Each part was measured twice by each operator. I don’t believe it is correct to run that Gage R&R study (crossed) again for this data.Do you have any suggestions on what to do? I would like to compare the two gages and show that one is more accurate and repeatable than the other.Thank you in advance for any help you can offer,~RUTH~Ruth D. AyalonMoog(716) 687 – [email protected]


    Bob J

    Hi Ruth!
    I noticed that you haven’t gotten a reply yet so will give it my 2 cents….
    The GR&R will give you a measure of how much variation is associated with the measurement process…  This is related to the gage precision….  To do this properly you need to run the classic GR&R using one measurement system (gage) and at least 2 people.  Since you have already completed one gage GR&R you should run the same evaluation on the second gage to be able to accurately contrast their performance….
    The second area is that of accuracy…  A GR&R will not help you here…  To accurately assess accuracy you need to look at how the measurement system (gage) performs against a known standard.  Most cal labs (inhouse or outhouse) should be able to verify the accuracy of the gage(s) in question…. 
    Since you are planning to contrast the accuracy and precision of both gages you also will want to use similar evaluation processes (people, samples etc) for both.
    Hope this helps…..
    Best Regards,
    Bob J


    Vivek Shrivastava

    Lets understand first what we are trying to do in GRR.
    we are trying to see what ratio of observed variation is being contributed by measurement system. And we want SS (measurement system) to be less than 30% of SST.
    Now if we can calculate SS (measurement) and SST our problem is solved.
    What I’ll suggest is measure your 10 parts by multiple operators, each part being measured by every operator multiple times. and every part to be measured by every operator on both the gages.
    Then do ANOVA wrt. part number. minitab reprt will give you SS part, SST and SS error. Divide SS error by SST and this is the contribution due to measurement system variation.



    Since the gage is the main piece of the puzzle, be sure to use the same operators and parts when completing both GR&R.  May sound trivial, but this will eliminate any question about the validity of this test.  I agree with the GR&R analysis and even ANOVA from the previous posts, but don’t forget about control charts.  A simple X-bar and R chart for each gage sure will tell you a lot and helps quantify your gage expectations when running later experiments.



    Mr.  bob j  is  absolutely  right.
    you need  minimum 2  people  for  gage  r&r
    then you may comapre the results of both the gages  in terms of
    1) %contribution
    2)%study variation
    3) distinct catagories.
    And then decide which measurement system is  best.
    Accuracy has nothing to do with G R&R.
    We have calibration  for  that.


    Thilanka Marasinghe

    Pls advice me how to get Gage R&R study results after entering the data 



    Hai Ruth,
    You can compare it in a lot of ways. Before you choose one: choose what you want to decide upon.
    Here my view (summarized)

    AbsoluteEvaluation: If you want to prove that each gage has a good Gage r&R value you just have to do the other one also (no shortcuts)

    RelativeComparison: Just fill in gage1 and gage2 in the operator column and perform your gage r&R. The reproducibility value you get is an indication how much different the 2 gage’s are FOR THAT ONE OPERATOR ONLY!. Mathematically not nice, but practical. Alternative: Regression between the 2 gage’s (scatterplot) or paired t-test for bias only.So yes since ‘all’ data concerns 1 operator you can only draw the conclusion on 1 operator.
    And I hope that the 2 operators are representative for all operators you get data from (because that’s your assumption for the conclusion). Note: Mathematically 2 operators is not enough. It should be >=30. But since that is totally unrealistic the ‘standard’ that everybody (should) use is 3 operators.
    hope this helps.


    Michael Mead

    Hello Ruth, it seems no-one is answering your question directly. Here is a procedure that will work:
    Calculate the range for each of your 20 pairs of measurements.
    Find tthe average range for each gage. (R1 and R2) Call R1 the larger range.
    Find F’ = R1/R2.
    If this ratio is greater than 1.9 then the gages have different amounts of variation with 95% probability.
    For the guys who always check what I write: Use Juran’s Quality Control Handbook, 4th edition, Appendix II, Table M. 
    Good luck.

Viewing 8 posts - 1 through 8 (of 8 total)

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