Destructive Gage RR - Using Crossed - want your thoughts

Hi, I am analyzing a Gage RR result, in a thermoforming process (plastic sheet run through a mold), and chose to select my part variation by selecting different mold cavities. I chose to have 10 parts, 3 ops, 2 repeats. The 10 parts (cavities) were selected from the mold that provided the most part variation. I chose cavities that produced the highest, lowest and in the middle range. I then made an assumption that each consecutive shot would represent a common part - ie, each cavity 3 for six consecutive shots represented the same part. So my repeats are created by consecutive cavities. I didn't select my parts until the machine had been stabilized (but recognize there could be some material variations). So, here is my question, since I am making the assumption that the 6 consecutive shots represent the same part (by cavity), I should be able to analyze the gage using a crossed method. By the way, the product is PET, and the test involves instron machines. Any thoughts or feedback? Thanks in advance.

The correct form of analysis would be a nested study. See my blog for a brief discussion of non-replicable R&R studies.

Hi Miner, I saw your posts, but I am not sure I agree. I think you can use a Crossed method, when the assumption is that you have created 'like' products (to represent the same part). The minitab statguide shows that as well. I think Nested is used when you can't set it up like that or make that assumption. Help me understand your thoughts on it. I too have been heavily involved with gage rrs and noticed you prefer only a nested analysis, but other searched have come up to similar thoughts of what I am suggested. Just want more insight. I also think using the different cavities is giving me my part to part.

I reviewed Minitab's StatGuide and found the guidance to which you referred. While I agree with the guidance in principle, you are applying it is a different mode from which it was intended. It could still work, but the assumptions depend heavily on whether the consecutive runs on a single cavity are indeed similar enough to claim that they are the same part. Do you have data that supports this assumption?

If I were running the analysis, I would run it using both approaches, compare the results and explore any differences. Since you appear to have Minitab and the effort involved in performing the Nested R&R is no different from the Crossed study, why are you reluctant to use it? The nested study is actually more sensitive than the crossed study if you assumptions are incorrect.

I don't think I have it set up like a nested design, and am not sure it makes sense in my design. Nested would assume that 2 consecutive cavities are the same for one operator, and if that is the case, I should just assume that all 6 consecutive are the same. These are high speed machines, so the least amount of variation will come with that scenario.

Unfortunately, I don't think I have the data, because this org has never done a gage RR; I'm new to the company and am bringing in opex tools and programs. Can't trust the data to see if it is the same anyway.

Doing one more post, so I can PM you

your assumption that sequential part from the same cavity of a high speed injection molding process will be almost identical is a good assumption.

as with any assumption, only the data will tell you the truth. I have used this very approach for similar - and exact - situations and have performed the study as a crossed study. (but I always check the graphical results and put less emphasis on the statistic that is calculated...). When my assumption is wrong it is clear in the graph and I redo the study - or sometimes jus the analysis as nested...

Miner's point is well taken: perform the study and the analysis both ways. then you will see...

Bev, thanks for more insight. The process is Thermoforming PET vs injection molding, but feel the assumption is valid. I did run it both ways, but I think I have it set up wrong in Mtb for the nested way, since I number each part the same (for each cavity). How would you do it otherwise?

Also, would like to hear your thoughts on what I should look at, at the results to determine if either is a better analysis.

That is the issue. In Minitab, select File > Open Worksheet > Look in Minitab Sample Data Folder > Select GAGENEST.MTW > Open then look at how the parts are numbered in the worksheet.

In your example, number parts for operator 1 as 1 - 10, for operator 2 as 11 - 20, and for operator 3 as 21 - 30.