What is the appropriate way to write a conclusion from a % contribution result?

[dzarr]

Hi,

I have just been introduced to MSA and I have a couple of questions.

I have been studying on ways how to mount accelerometers onto human skin in order to measure skin accelerations (vibration). I was not happy with the result as the acceleration amplitude tends to fluctuate during repeated measures.

As I was searching for a way to analyze measurement capability, I stumbled upon AIAG MSA 4th Ed. After having improved the method and the repeatability on a single operator I have moved on to the factorial analysis using the recommended 10 parts, 3 operator and 3 replicates.

I have repeated my test and increased appraisers number from 3 to 6 to narrow down the confidence interval for %StudVar and ndc. I have since moved away from %StudVar to %Contribution which I think is the more appropriate metric in my situation. I should say I am happy with the outcome, the upper limit of gage r&r 95% CI is below 9. Attached herein is the minitab two-way ANOVA gage study result with 95% CI.

As I have had no formal training on Gage Analysis (mostly reading) I am concern that I might have got to a wrong footing.

My questions;
1. Is it statistically correct to make the following conclusion from the analysis data.

The measurement error contributes only 2.92% from the total measurement variance. As the upper limit at 95% CI is 6.51% which is less than 9%, this is an acceptable method for measuring.....

2. I wanted to cite the reference for the %contribution. GRR<1, 1>=GRR<=9, GRR>9. AIAG only provided the guidelines on %StudVar and ndc. Who first recommended the limits for %contribution? one which I could cite.

Thanks

A.Dzarr

 

[Stijloor]

A Quick Bump!

Can someone help??

Thank you very much!!

 

[dzarr]

I wonder if anyone who has been following the history of Gauge R&R may have the knowledge as to when %contribution came into the picture. I suppose it should be around the time when ANOVA came into prominence in regards to Gauge R&R. I had difficulty zeroing in on this info which could help lead me to the answer to my second question. Thanks

 

[Miner]

My questions;
1. Is it statistically correct to make the following conclusion from the analysis data.

The measurement error contributes only 2.92% from the total measurement variance. As the upper limit at 95% CI is 6.51% which is less than 9%, this is an acceptable method for measuring.....

This is actually more statistically correct than any of the other metrics as it is based on the variance, which is additive and therefore can legitimately be used in a ratio or percentage. The other metrics are based on the standard deviation, which is not additive and therefore cannot (but obviously is) be legitimately used in a ratio or percentage.

2. I wanted to cite the reference for the %contribution. GRR<1, 1>=GRR<=9, GRR>9. AIAG only provided the guidelines on %StudVar and ndc. Who first recommended the limits for %contribution? one which I could cite.

I have only seen guidelines for % contribution through Six Sigma training classes. I have never seen an actual citation as to where they came from.

 

[Bev D]

I am not aware of any citation for this either. I have read different approaches to deciding on what is acceptable and most are roughly equivalent to eh old rule of thumb about measurement resolution. certainly there is no mathematically precise derivation - only popular convention. much like the 5% alpha choice...