The following is a very interesting article by Don Wheeler, a highly respected expert in SPC and data analysis. Don expresses his opinions of the suitability of traditional Gage R&R versus Intraclass Correlations, and backs it up with simulations.

Good Data, Bad Data, and Process Behavior Charts (http://www.spcpress.com/ink_pdfs/Good%20Data,%20Bad%20Data.pdf)

Those of you that are not forced into using Gage R&R by your customer may want to investigate further.

I highly recommend that everyone read Dr. Wheelers book, "Evaluating the Measurement Process". I really feel the EMP process is superior to standard GRR. :D :D

I highly recommend that everyone read Dr. Wheelers book, "Evaluating the Measurement Process". I really feel the EMP process is superior to standard GRR. :D :D

Don Wheeler has published a new book, EMP III Using Imperfect Data". At 316 pages, it is a complete overhaul of the previous EMP manual.

I realize that this thread is more than a year old, but I thought there may be interest in revisiting the topic. What are your opinions regarding AIAG GR&R compared to Wheeler's EMP methods? I just received this article from Wheeler's e-newsletter this morning. I am going to engage our customers in this discussion to understand why they are specifying the AIAG methods. I'm not a "stats brat" so I don't have insight to pros and cons. Wheeler seems rather adamant that the traditional GR&R study is not correct and that those who use it are fooling themselves. Any thoughts or comments are appreciated!

I realize that this thread is more than a year old, but I thought there may be interest in revisiting the topic. What are your opinions regarding AIAG GR&R compared to Wheeler's EMP methods? I just received this article from Wheeler's e-newsletter this morning. I am going to engage our customers in this discussion to understand why they are specifying the AIAG methods. I'm not a "stats brat" so I don't have insight to pros and cons. Wheeler seems rather adamant that the traditional GR&R study is not correct and that those who use it are fooling themselves. Any thoughts or comments are appreciated!

There's a contemporary discussion thread (http://elsmar.com/Forums/showthread.php?t=25541) on this subject.

Thankfully, I don't get involved with GR&R here on my job. I would definitely agree with Dr. Wheeler's assertion that ratios of standard deviations are meaningless. Consider, for example, Design of Experiments and ANOVA using the F test. Everything there is based upon ratios of variances, not standard deviations.

I don't have a copy of the AIAG manual with me right now, but I think the manual does calculate the variance ratios. Is it ok if GRR acceptance criteria is changed and based on variance ratios of 1%, 4% and 9% instead of standard deviations ratios of 10%, 20% and 30% ? What would be the difference in terms of interpretation of physical contributions of error terms?

I love all these, thanks to all of you!