iSixSigma

Comparison of Measurement Methods

Six Sigma – iSixSigma Forums Old Forums General Comparison of Measurement Methods

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #44212

    Lorax
    Participant

    Folks,
     
    Another tricky one I need your help with:
     
    I’m trying to compare three candidate test/measurement methods in order to identify the one which consistently returns the value closest to the “real” dimension.
    I have 4 columns of measurements taken from 120 items.
     The first column is the “real” value (which is a 1 or a 0)
    The other three are the values returned from each of the candidate methods when they measured that same item and are continuously variable between 0 and 100
     
    What’s the best way of spotting the best test method? A method which shows some nice pictures as to why would be a good thing (this has to be explained to some of our scientific brethren).
     
    Many thanks
     
    Lorax

    0
    #141412

    Ang
    Participant

    I guess you could calculate the variances, and see which candidate has the lowest variance, since this is the candidate whose answers are closest to the ‘real’ value (or less far off as the other ones).
    Peter

    0
    #141742

    melvin
    Participant

    Not wanting to replicate the effort of others, I suggest you go to the minitab website & search on attribute measurement systems analysis.
    I think that attribute agreement analysis will help you here.
    Bob

    0
    #141743

    boettler
    Member

    Try to have the ffg.:1. Boxplot – variability chart so that each appraiser will see their variation with others2. ANOVA – to see if the different readings are significant3. Post Hoc test – Tukey Kramer – to see who among the appraisers have the closest significant readings4. GRnR – to determine the gare repeatability and reproducibility and determine also the ndc (number of distinct categories)

    0
    #141757

    villageidiot
    Member

    How are you defining “best”?  It appears to be fairly simple to compare the accuracy of each method (%) as well as its precision (std dev).  Now use Coefficient of Variance (stddev/mean) to determine variability. 
    Double check this, as I am making most of this up as I go along.  Good Luck.

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

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