iSixSigma

How Do We Ensure Proper Sampling for Gage R&R?

Six Sigma – iSixSigma Forums General Forums Methodology How Do We Ensure Proper Sampling for Gage R&R?

This topic contains 5 replies, has 5 voices, and was last updated by  Chris Seider 8 months, 3 weeks ago.

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #208239

    Gorur Sridhar
    Participant

    An “exact” sampling – while not practical in real life, this pulls parts corresponding to the 5th, 15th, 25, …, and 95th percentiles of the underlying normal distribution and forms a (nearly) “exact” normal distribution as a means of seeing how much the randomness of sampling affects our estimates.

    Parts are selected uniformly (at equal intervals) across a typical range of parts seen in production (from the 5th to the 95th percentile).

    Parts are selected uniformly (at equal intervals) across the range of the specs, in this case assuming the process is centered with a Ppk of 1.

    8 of the 10 parts are selected randomly, and then one part each is used that lies one-half of a standard deviation outside of the specs.

    …The question is “unless we measure the parts how do we ensure that we are selecting the parts at the extreme of the specs, outside the specs. etc”

    It is the “Chicken or the Egg situation” – Please comment.

    0
    #208241

    Mike Carnell
    Participant

    I don’t understand what you want to accomplish. First let’s be clear in your procedure you are not selecting randomly. So what do you want? An estimate of the mean? Std dev? Distribution?

    It sounds like you want such perfect knowledge that you are going to have to do 100% sampling (The infamous 100% sampling which is sorting). Have looked into confidence intervals. I am not sure but it seems like that may be a part of the picture you are missing.

    0
    #208242

    Gorur Sridhar
    Participant

    Hi Mike,

    Selecting parts for conducting an effective gage R&R requires random samples and at the extreme values of the tolerances.
    The above can be selected only after the R&R is known. So the issue is ‘Which comes first’. I hope I am clearer now.

    0
    #208285

    Strayer
    Participant

    It sounds like you might be conflating GR&R with calibration. They’re related but different.

    0
    #208373

    David Hampton
    Participant

    Hi Gorur
    You’ve been given poor guidance here I’m afraid.
    The best way to get a representative sample is to make a random selection over a reasonably long period of time. DO NOT then distort the sample by adding in an unusually large and an unusually small part. This will inflate the std dev of the parts in the sample and thus cause GR&R to be calculated incorrectly (appear to be lower than it really is). If you are reporting the GR&R results to your customer based on this, it amounts to fraud.

    Getting a represenative sample is challenging and there is a much better way, which is to get QC inspection records from dimension in question and get a bunch of measurements from them – say, a week or a months’ worth of data – and calculate the standard deviation of these measurements. Then use that as the denominator in the GR&R calculation rather than the SD of the parts in the sample of 10. Much more reliable and robust way of doing it (and highly recommended by Minitab as well I might add)

    cheers

    David

    0
    #208498

    Chris Seider
    Participant

    Unusually large is a “wide term”. You want to make sure you can measure marginal or even those out of spec level but should still be typical process results. However, if you know the process, you’ll know how to balance the approaches.

    0
Viewing 6 posts - 1 through 6 (of 6 total)

You must be logged in to reply to this topic.