The error comes from the fact that the difference between the averages of your operators is unusually small, resulting in the square root of a negative number in the calculation of the AV term. Calculation of AV is defined by the AIAG MSA manual, which I can't avoid. The solution is fairly simple. The latest version of my GR&R spreadsheet performs a complete Analysis of Variance (ANOVA) and avoids the negative square root problem. Please refer to the numerical summary tab for this. If you have any additional comments or questions, please feel free to e-mail me at dean@CpkInfo.com
--Dean Christolear, 09-24-2011

Many thanks.
--Kennie, 03-03-2011

Recently used the GRR spreadsheet and found an issue in the calcualtions for Appraiser variation. A #Num! error occured as a result os the calcualtion trying to take the squareroot of a negative number. See Below
AV = SQRT{(Xbar diff x K 2)^2 - (EV^2/[(# parts) x (# trials)])}
Xbar diff = 0.030933333
K2 = 0.5231
#parts = 5
# trials = 3

AV = SQRT(0.000262 - 0.00189)
AV = #NUM!

When a negative number occurs shouldn't the AV value Zero out?

If we look at the Anova method the Reproduceability shows at Zero as well as the interaction with the part. This is due to the p value exceeding the alpha of .25 indicating the appraiser variation is not significant. When this occurs you may combine the appraiser variation(Reproduciblity) with Repeatability.

Also consider adding one more Line Chart that I have found to be wonderful. A Gage Performance Curve showing Y Axis Pa (Probability of Acceptance), X Axis the Measurement / Dimension.
Probablity Lines:
Pa
Part Variarion
Operator error
Gage error
Total error.

--Don Carnley, 03-22-2010