[FONT="]So here are the results of the GRR (and a bit of a rant!*Disclaimer*)[/FONT]
[FONT="]Two-Way ANOVA Table With Interaction [/FONT]
[FONT="] [/FONT]
[FONT="]Source DF SS MS F P[/FONT]
[FONT="]Part no 74 0.0226890 0.0003066 2.5508 0.000[/FONT]
[FONT="]Operator 3 0.0059907 0.0019969 16.6131 0.000[/FONT]
[FONT="]Part no * Operator 222 0.0266843 0.0001202 1.1980 0.073[/FONT]
[FONT="]Repeatability 300 0.0301000 0.0001003[/FONT]
[FONT="]Total 599 0.0854640[/FONT]
[FONT="] [/FONT]
[FONT="] [/FONT]
[FONT="]Alpha to remove interaction term = 0.25[/FONT]
[FONT="] [/FONT]
[FONT="] [/FONT]
[FONT="]Gage R&R [/FONT]
[FONT="] [/FONT]
[FONT="] %Contribution[/FONT]
[FONT="]Source VarComp (of VarComp)[/FONT]
[FONT="]Total Gage R&R 0.0001228 84.05[/FONT]
[FONT="] Repeatability 0.0001003 68.68[/FONT]
[FONT="] Reproducibility 0.0000224 15.36[/FONT]
[FONT="] Operator 0.0000125 8.56[/FONT]
[FONT="] Operator*Part no 0.0000099 6.80[/FONT]
[FONT="]Part-To-Part 0.0000233 15.95[/FONT]
[FONT="]Total Variation 0.0001461 100.00[/FONT]
[FONT="] [/FONT]
[FONT="] [/FONT]
[FONT="] Study Var %Study Var[/FONT]
[FONT="]Source StdDev (SD) (6 * SD) (%SV)[/FONT]
[FONT="]Total Gage R&R 0.0110805 0.0664831 91.68[/FONT]
[FONT="] Repeatability 0.0100167 0.0600999 82.88[/FONT]
[FONT="] Reproducibility 0.0047376 0.0284253 39.20[/FONT]
[FONT="] Operator 0.0035371 0.0212228 29.27[/FONT]
[FONT="] Operator*Part no 0.0031517 0.0189102 26.08[/FONT]
[FONT="]Part-To-Part 0.0048271 0.0289627 39.94[/FONT]
[FONT="]Total Variation 0.0120863 0.0725178 100.00[/FONT]
[FONT="] [/FONT]
[FONT="] [/FONT]
[FONT="]Number of Distinct Categories = 1[/FONT]
[FONT="]Turns out the reason the Results were different in the 1st place is because the operators didn't measure the parts correctly on my side.
These parts are from a new supplier so they were meant to perform the measurements in the same way as they did the parts from the old supplier. I assumed that they had been trained to do this correctly...how wrong was I! I found this out when I sent them the protocol for the GRR.[/FONT]
[FONT="]So once I'd re-trained them we performed the GRR and the results were equivalent. There was a minor discrepancy, however... [/FONT]
[FONT="]As you can see the caliper method is inadequate. The method is as follows: measure the wall thickness at 4 points 2 in the X plane and 2 in the Y plane, i.e. at 12, 3, 6 & 9 on a clock face. Calculate the average and report that. My operators had only been taking 1 value. I was going to see whether, if they took the measurements in the same tube orientation everytime would the variation decrease.[/FONT]
[FONT="]I started to think up a 2nd quick study to see if I could decrease the gage variation when I notice the tubes have 4 points marked on the tip. Knowing full well what they'd done I asked why they'd marked the tubes. "Well we realised that if we both took the 4 measurements in different places it might increase the variation so we marked them all up and then both measured them at the same point" therefore artificially decreasing the variation in the operator reproducibility...
[/FONT]
[FONT="]Moral of the story: Just because people are highly educated does not mean they have an ounce of sense! Watch them at all times! [/FONT]
[FONT="]The supplier maintains that they don't have any issues with any other customers when they use this technique so we should look at widening the tolerances (I think this is a bit of a joke but nevertheless I won't be signing off on it!)[/FONT]