iSixSigma

GRR study

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #37702

    Sumit Banerjee
    Member

    Hello
    I am a newcomer in Six Sigma. I am working in a Ferro-Chrome production unit as Black Belt. I aim at conducting a GR&R study of our chemical analysis. Each of the three chemists were asked to conduct analysis of samples from three furnaces, twice. As the product is very heterogeneous in nature (it even varies from sample to sample of a particular furnace), I took utmost care in sample preparation. Please guide if GR&R (Nested) is a more appropriate tool in this situation than GR&R (Crossed).
    Regards
    Sumit

    0
    #111471

    Bob J
    Participant

    Sumit,
    Since the analysis is destructive in nature you should use the GR&R (Nested) for your study…
    Best Regards,
    Bob J

    0
    #111563

    Marc Thys
    Participant

    If the same sample is tested several times by different people than it is, per definition, a “Crossed” GR&R.
    You only have a nested design when each sample is tested only by one operator (by necessity or by design).
    A crossed study is preferrable because you get more information from it.

    0
    #111617

    Bob J
    Participant

    Sumit,
    Note that both answers (Marc and myself) hinge on whether the sample testing is destructive in nature.  If the same exact sample can be retested over and over again, then Marc is right and a crossed GR&R is your best choice.  If the tested sample is destroyed by the test process and another sample must be used for additional test, then a nested GR&R is your best choice…
    Hope this helps….
    Best Regards,
    Bob J
     

    0
    #111633

    Ron
    Member

    If all operators measure parts from each batch, then use Gage R&R Study (Crossed). If each batch is only measured by a single operator, then you must use Gage R&R Study (Nested). In fact, whenever operators measure unique parts, you have a nested design.

    0
    #111651

    mjones
    Participant

    If you assume this is a “destructive test” and that each sample cannot be measured by another person, you have a nested design. One view is this is certainly true since a chemical test typically does destroy the sample. But, if you “accept” this assumption absolutely, you cannot measure repeatability because an inspector cannot actually measure the same sample twice, since the test destroys the sample. Therefore, in destructive Ga R&Rs, we use samples that are as similar to each other as possible and have the same person measure them to check repeatability.
    Usually, if I am able to get samples that are “very similar” so I can check repeatability, I can usually get a few more samples that are also “very similar” and do a crossed Ga R&R; which is much more powerful — as stated in a previous post.
    Practically speaking, if we sample from a lot of homogeneous product, we can usually assume all samples are essentially the same so we can perform a valid crossed Ga R&R.
    In the original post it was stated that there is a great deal of sample-to-sample variation. If so, my suggestion to assume homogeneous samples may not be valid. Or, it may be that you actually have a problem with sampling, or testing, or product homogeneity, or repeatability, or reproducibility, or some combination of these. And/or, it may be that products you are running the Ga R&R on are so consistent, relatively speaking, that your system is simply not capable of detecting differences in them.
    I suggest you take multiple samples from each lot (lots that represent a wide range of values) and perform multiple tests by operators and with multiple operators. First, presume homogeneity of samples (and attempt to obtain them in a way to be as consistent as possible), run as crossed and look for issues of interaction and sources of variation. If this does not work, you can go to Destructive Ga R&R and, rerun the analysis. Or, rerun it as crossed with substantially different lots. h

    0
    #111676

    Sumit Banerjee
    Member

    Dear Bob & Marc
    Thank you for your suggestions. What I think that for any chemical analysis, samples may vary but cannot be destructive in nature. So I have tried GR&R crossed which came out to be 21%.
    But GR&R nested is only 7.7.
    What is your comment on it.
    Regards
    Sumit

    0
    #111683

    Bob J
    Participant

    Sumit,
    It’s due to sample to sample variation with is impacting the variation in the crossed study…  See mjones’s excellent post…
    Best Regards,
    Bob J

    0
Viewing 8 posts - 1 through 8 (of 8 total)

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