The meaning of %StudyVar opposed to %Contribution

R

Reach

Hi guys,
I'm doing some research and trying to understand the meaning of numbers resulted from Gage R&R.

I got an example set of data for a Gage R&R that shows following results:
----------------------------------------------------------------------
Two-Way ANOVA Table With Interaction

Source DF SS MS F P
Part 2 0.0001881 0.0000941 397.099 0.000
Operator 2 0.0000017 0.0000008 3.526 0.131
Part * Operator 4 0.0000009 0.0000002 2.520 0.077
Repeatability 18 0.0000017 0.0000001
Total 26 0.0001925

Alpha to remove interaction term = 0.25

Gage R&R

%Contribution
Source VarComp (of VarComp)
Total Gage R&R 0.0000002 1.96
Repeatability 0.0000001 0.88
Reproducibility 0.0000001 1.07
Operator 0.0000001 0.63
Operator*Part 0.0000000 0.45
Part-To-Part 0.0000104 98.04
Total Variation 0.0000106 100.00


Study Var %Study Var
Source StdDev (SD) (6 * SD) (%SV)
Total Gage R&R 0.0004562 0.0027374 13.99
Repeatability 0.0003066 0.0018397 9.40
Reproducibility 0.0003378 0.0020269 10.36
Operator 0.0002579 0.0015472 7.91
Operator*Part 0.0002182 0.0013094 6.69
Part-To-Part 0.0032290 0.0193739 99.02
Total Variation 0.0032611 0.0195663 100.00


Number of Distinct Categories = 9
----------------------------------------------------------------------

I have found equations and was able to do manual mathematics to obtain those results. I understand (or at least i think i do) why %contributions add up to 100% and how that number is used (or explain the meaning of that number to layperson).

However, although i can follow how to calculate those numbers, I don't understand why get the second set of numbers with SD's and %Study Var's, and how they are used.

I know the criteria for %contribution, and %StudyVar (i asked you guys for the help on that as well last week or so :D ), but i'm now trying to understand actual meaning of them.

I guess the question that confuses me the most is the use of percentages (%) for study variations. It seems natural when we find percentages of certain things, we want that to add up to 100%.
Again, i understand they dont add up to 100 because we're using SD's but don't understand why we use it.

I'm sorry i'm rambling on and my wording is really confusing. I guess i'm not exactly sure what to ask since i recently started studying about gage r&r and only have very crude knowledge about the test.

Thanks!!!


ps. sorry for the messy data; i'm new to pasting data, and posting thread skills><
 

Bev D

Heretical Statistician
Leader
Super Moderator
well the % study variation is off because you can't* divide standard deviations. you have to work with variances.

*of course you do anything you want it's just not mathematically correct.

% contribution when resulting from a properly calculated ANOVA uses the variances and properly deconstructs the contribution of each factor.
 
R

Reach

hey
thx for the reply, but i think there's a little misunderstanding between you and me (probably because my poor composition of writing>< )

i haven't done any math, and those results are simply taken out of minitab.
i'm wondering, what does the %StudyVar and SD's mean? and why does minitab (or any other programs) calculate those if we can check %contribution to see if the gage is proper for measuring. (if less than 1% or whatever maybe)
 

Bev D

Heretical Statistician
Leader
Super Moderator
Yes sorry. Let me clarify as well: I know *you* didn't do that math; Minitab did. I don't know why the software does that. It is misleading because of the mathematical incorrectness of dividing SDs. Their bad. Use % contribution. %Study is not value add.
 

Miner

Forum Moderator
Leader
Admin
It is not really Minitab's fault about the % Study Variation. They, as well as all the other MSA software companies, are just copying the AIAG MSA manual, which, in turn, copied the SD error from the % Tolerance approach, which I first saw in a General Motors supplier publication.

The methodology is flawed, and makes a gage appear worse than reality, as demonstrated by several respected sources, including Donald Wheeler (See "An Honest Gauge Study").

If you are not required by the customer to follow the AIAG MSA manual, Wheeler's approach is mathematically sound, and is a better approach. However, customers that are familiar with MSA are probably tuned into the AIAG approach and will probably resist a different method.

The AIAG method is overly conservative, meaning a gauge that passes is definitely good, but a gauge that fails is not necessarily bad.
 

Bev D

Heretical Statistician
Leader
Super Moderator
It is not really Minitab's fault about the % Study Variation. They, as well as all the other MSA software companies, are just copying the AIAG MSA manual, which, in turn, copied the SD error from the % Tolerance approach, which I first saw in a General Motors supplier publication.

yes it's amazing how an error can be replicated over and over and all of a sudden it becomes 'the law' even though it's been disproved many times over and almost no one likes it. Teh origin - as near as I can tell - is the article "The rubber ruler" by Robert Traver published in either 1959 or 1960...

It's a shame because it perpetuates poor practice - you would think ASQ would do something about it - at least their mision statement says they should...oops.
 

Miner

Forum Moderator
Leader
Admin
Teh origin - as near as I can tell - is the article "The rubber ruler" by Robert Traver published in either 1959 or 1960...

I located it for sale in the ASQ archives. It is dated 1962, and Robert Travers was working for General Electric. A different General appears to be the culprit.
 
L

Licht

Hi.:bigwave:

I have a question: If the %StudyVar does not add up 100%, what's the meaning of the percentage value ? For %Contribution I understand the meaning because I have the notion of greatness.

For example, in the picture attached, what's the meaning of the value 27.86% for Total Gage RR ? If it does not add up 100% I can't say that the Total Gage RR represents 27.86% of the Total Variation Study. Right ?
It's not clear to me... :confused:

regards.
 

Attachments

  • gage.png
    gage.png
    6.9 KB · Views: 245

Miner

Forum Moderator
Leader
Admin
This is fairly straightforward. % contribution is based on variances (s^2). Variances are additive and add up to 100%. % SV is based on standard deviations (s). Standard deviations are not additive and do not add up to 100%.

This is one of the fallacies of the AIAG method. See Dr Wheeler's approach for a better explanation.
 
Top Bottom