Dear all.
This is Mike.
(sorry for language is not good)
I am having problem with trying gage r&r
1. I had tried to find what will happen if I didn't use random but comfort orderly manner.
-->Every data I have is conducted at orderly manner, but these are only data available.
-->I had tried to find what happen when I use non-random data, but found only ambiguous answer which assume there will be more bias at conclusion. Which is not useful for me.
*Here is my data concept. (all but Y is nominal number)
day (1,2)
operator (1,2,3)
product lot # (1,2,3,4)<==(I personally think that including several lots and sampling point in the gage R&R is good because these will reflect the variation in overall process)
sampling point (1,2,3,4,5)
and all data are so complicate(crossed and nested and both etc.) since data are mined from everywhere.
It is like, one lot has sampling point of 1,2,3 but an other lot has 4,5 and, Day 1 has two operator but Day 2 has only one operator...
From above, each operator was given needed vial.
But unlike usual crossed design, test is quit destructive so that identical vial can not be re-used-by-the-other-operator
Important thing is, One operator can re-test a vial several times, set to 3 times.
2. How can I handle this data set ?
I had tried crossed, nested, crossed then nested,
I was so confused how can I assign the factors.
3. I found that if I use X with all of above (day-oper-lot-samp) and X with only one (operator, with neglecting the other factors as if they were not assigned at the beginning of the experiment), the latter conclude much more satisfying R&R score. Former: ~95%, Latter: ~35%
I really have no idea why this happen. As per my common sense, the only difference is I gave 'more' information to the JMP in former case.
Thus I expected that more analytical information in JMP for former case but, it gives me confusing answer.
4. For me, I am using ELISA method, but it is assumed that my question is understandable for many others using 'Optical measuring method'.
The question is,
Required repeat in gage R&R is conducted as below.
-From a vial, 3 aliquots (a portion) were extracted.
-Each aliquot was filled in a well (you may have seen 96 wells micro-plate which analyst uses at biochemistry)
-So, filled 3 wells are now 3 repeat for me.
-This machine measures entire 96 wells at once
-I am now worrying that if I didn't satisfy minimal (if authority requires any) time-gap between the repeat.
Can I regard this 3 replicate (could I call this as replicate? not sure) as the required repeat in gage R&R. Could I do this?
5. For a reason I wanted to give more data to JMP, I defined Y as 'individual measured value', however, in the industry, we use average of 3 replicates.
Can I report this R&R based on the individual value though it doesn't equal to the usual value we use?
Thank you very much for reading all this far.
Mike.
