iSixSigma

Attribute Rank measurement systems evaluation

Six Sigma – iSixSigma Forums Old Forums General Attribute Rank measurement systems evaluation

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
    Posts
  • #37107

    David Martinez
    Participant

    I´m investigating about how to evaluate attribute-rank measurement systems.
    AIAG-MSA manual defines just how to evaluate attribute-binomial type measurement systems (go/no-go), but I would like to evaluate Multinomial measurement system.
    I believe that Likert Scale could be useful for this purpose. But I´m not shure if using a multinomial P.D as Normal probability distribution could be statisticaly correct. Or do I have to use Non-parametric test to evaluate this kind of measurement system? Do you know some article related to this issue?
    Thanks
    David Martinez

    0
    #108835

    Markert
    Participant

    David
    The AIAG manual gives tools for gauge R&R studies for attributes where the attribute is a go/no-go characterisation of a feature which could be measured as a variable. These tools should not be used for attibutes which only exist as attributes (e.g cracked / not-cracked) etc.
    Non-parametric methods based on ranks could be used, but only when the ranking structure exists within the frame of the experiment only, and also represents something which could be measured instead (even if the cost of doing so would be prohibitive). This is equivalent to performing ANOVA with a “rank-order transformation.”
    The use of a “Likart” appears to makes reference to a scale outside the frame of objects included in the study. For this reason it is incorrect.
    I would suggest you take a look at the differences between Nominal, Ordinal, Interval, and Ratio scales of measurement. This should help you work out what you can and cannot do with different types of measurement.
    Hope this helps.

    0
    #108954

    Perryman
    Participant

    Hi David,
    What is it you are trying to show? That the measurement system is effective in terms of predicting a result or in terms of consistency of output?  I’m not sure what ranking system you’ve used – have you looked at Multiple Attribute Utility Theories such as BCG or AHP, or Smart.  These are pretty effective.
    If your intent was to measure whether, for example the outcome of one of these ranking methods was indeed the best, it would imply that you want to compare the expected (predicted) result with the actual result – you could maybe consider Chi Square.  If, on the other hand, you want to compare the ranking results among groups or between groups – perhaps a GR&R is the way to go…
    Food for thought.
     

    0
    #108965

    Edwards
    Participant

    Path/Phil:
    Thanks in advance for your answers.
    In fact, I´m interested in evaluating of a Multiple-Attribute Measurement System, but I would like to estimate in a separate way Repetibility and Reproducibility errors, as R&R studies are doing with continuous variable, although I know, Multiple-Attribute-Rank Measuremente System are represented by Multinomial-distribution.
    Which is the correct approach to do this?
    Are there some articles or books regarding this issue I can read to get more information?
    Thanks again
    David

    0
Viewing 4 posts - 1 through 4 (of 4 total)

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