iSixSigma

Measurement system analysis

Six Sigma – iSixSigma Forums Old Forums General Measurement system analysis

This topic contains 4 replies, has 3 voices, and was last updated by  Sinnicks 19 years, 1 month ago.

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #26911

    M. Gerusky
    Participant

    There are a few methods you could use:
    Assuming you use Inspection capability:
    Effectiveness should be > 90%
    Probability of a Miss should be < 5%
    Probability of a False Alarm should be <5%

    If you choose to use Degree of Agreement:
    Repeatability should be > 90%
    Reproducibility should be > 90%

    0
    #65414

    Brian Terau
    Participant

    Those are some good figures to keep in mind. I agree with you. Thanks for posting.

    Brian

    0
    #65418

    Sinnicks
    Participant

    M. Gerusky — Where did you come up with the the figures? Why is 90% used for effectiveness. Shouldn’t it be closer to 99.9996 to meet six sigma?

    Doesn’t effectiveness take into account a miss? Why is there separate numbers?

    Thanks,
    Mark

    0
    #65419

    M. Gerusky
    Participant

    Mark-
    The topic is Measurement System Analysis (Gage R&R) with attribute data, not manufacturing capabilities (to which the lofty measure of six sigma applies).
    Gage R&R’s for variable data are fairly straightforward whether using Xbar&R or ANOVA. The AIAG has provided guidelines as:
    Ideal 10% or less
    Acceptable 10-20%
    Marginal 20-30%.
    Evaluating a measurement system for discrete data becomes more difficult. The two methods that I’m familiar with using attribute data are: Inspection capability and Degree of Agreement. The method of inspection cpability will yield three separte results:
    Effectiveness, probability of a false alarm and probability of a miss. (If you’d like the methodology I’d be happy to share it).
    Degree of Agreement is used when the “truth” of the samples is unknown. This will yield two separate measures: Repeatability and Reproducibility.
    Having the experience of conducting numererous studies with each of those methods, a result of over 99% is unseen and highly unlikely.

    0
    #65421

    Sinnicks
    Participant

    Thanks. That does make sense. I appreciate your help.

    Mark

    0
Viewing 5 posts - 1 through 5 (of 5 total)

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