L
Laura O
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
Thanks. here is the file
Thanks. here is the file
% of Contribution or % of Study Variation? Gage R&R using Minitab
B
Barbara B
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
The GRR% compared to the variation of parts in your study is 79.77%, so you can't distinguish between the parts and the number of distinct categories is 1: statistically all parts are identical. That's bad if you want to monitor your process (e.g. control charts) or evaluate process capability (e.g. Pp/Ppk).
Sometimes this high values occur due to a part selection which doesn't reflect the whole process range. There should be used 10 different parts for a Gage R&R. To check whether you have a measurement system issue or a part selection issue you could calculate the GRR% using a historical standard deviation (of the process).
If the GRR% (%process) values are similar to 80%, the measurement system uncertainty is too high to reflect process changes. If the GRR% (%process) is significantly smaller than GRR% (%study var) the selcted parts are too similar and the Gage study should be conducted with new parts which are different.
Since the tolerance width is much broader than the part variation the GRR% (%Tol) is acceptable with 6.1% < 10%. You can distinguish between good parts and bad parts and use this measurement system for part selection.
The GRR% compared to the variation of parts in your study is 79.77%, so you can't distinguish between the parts and the number of distinct categories is 1: statistically all parts are identical. That's bad if you want to monitor your process (e.g. control charts) or evaluate process capability (e.g. Pp/Ppk).
Sometimes this high values occur due to a part selection which doesn't reflect the whole process range. There should be used 10 different parts for a Gage R&R. To check whether you have a measurement system issue or a part selection issue you could calculate the GRR% using a historical standard deviation (of the process).
If the GRR% (%process) values are similar to 80%, the measurement system uncertainty is too high to reflect process changes. If the GRR% (%process) is significantly smaller than GRR% (%study var) the selcted parts are too similar and the Gage study should be conducted with new parts which are different.
Since the tolerance width is much broader than the part variation the GRR% (%Tol) is acceptable with 6.1% < 10%. You can distinguish between good parts and bad parts and use this measurement system for part selection.
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
I would also add that the study involves photometric data, which is a notoriously difficult thing to corral for MSA. It appears that what the data is actually saying is that a significantly different approach is needed.
I would also add that the study involves photometric data, which is a notoriously difficult thing to corral for MSA. It appears that what the data is actually saying is that a significantly different approach is needed.
Last edited:
B
Barbara B
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
What method would you recommend to assure that the values from this kind of measurement system are reliable?
Excellent point
What method would you recommend to assure that the values from this kind of measurement system are reliable?
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
It's not easy to say without knowing more about the things being measured (how many, geometry, color etc.), the instrument itself, and operator training. There are lots of variables.
ETA: When it comes to MSA, we need to be less focused on pass/fail and more attuned to what the data is saying. A GR&R study that tells you that something is significantly wrong is not a bad thing.
Excellent point
It's not easy to say without knowing more about the things being measured (how many, geometry, color etc.), the instrument itself, and operator training. There are lots of variables.
ETA: When it comes to MSA, we need to be less focused on pass/fail and more attuned to what the data is saying. A GR&R study that tells you that something is significantly wrong is not a bad thing.
L
Laura O
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
Barabara and Jim-
Thanks for your help. I do PPAP Approval and the document you saw was submitted by the supplier. It's up to us to determine if it passes, so that we can approve the PPAP. I thought I knew how to read these reports but this one does not indicate pass/marginal/fail. I am not just about PASS/FAIL, I also want to understand the data, but Pass/Fail is a requirement of PPAP.
I have requested the raw data, and the other items such as instrument, amounts and such. As with all of our suppliers, we require that the GR&R is done using the full range of tolerance allowed.
The material is rolled goods - skin. When I saw the 79% I thought I must be reading it wrong. But I have also had submissions where the supplier was confident that either A. We wouldn't notice or B. They didn't understand it (or care).
I appreciate your feedback and it's nice to know that it's not that easy to dissect.
Thank you both again for your quick and helpful response.
Barabara and Jim-
Thanks for your help. I do PPAP Approval and the document you saw was submitted by the supplier. It's up to us to determine if it passes, so that we can approve the PPAP. I thought I knew how to read these reports but this one does not indicate pass/marginal/fail. I am not just about PASS/FAIL, I also want to understand the data, but Pass/Fail is a requirement of PPAP.
I have requested the raw data, and the other items such as instrument, amounts and such. As with all of our suppliers, we require that the GR&R is done using the full range of tolerance allowed.
The material is rolled goods - skin. When I saw the 79% I thought I must be reading it wrong. But I have also had submissions where the supplier was confident that either A. We wouldn't notice or B. They didn't understand it (or care).
I appreciate your feedback and it's nice to know that it's not that easy to dissect.
Thank you both again for your quick and helpful response.
A
Allattar
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
Be careful with taking full range of tolerance measuremesnts. Ideally you should have sample that represent the process being measured.
If not try using an historical standard deviation to get %Process.
If the samples do not represent the process that will be measured by the gage then the %Study variable cannot be applied to measuring the process. Basically %study variable is can I measure the samples in the study.
If you take samples to cover the tolerance width you can increase the variation of the samples to greater than the process.
For example. Take a process with a mean of 0 and standard deviation of 1. The worst possible scenarios for sample collection would be either all parts are 0, or you have 5 at -3, and 5 at +3.
The population has a sd of 1. In the scenario of 5 samples at -3, and five at +3, you will get a standard deviationof over 3.
A very important question to ask is do the samples they have measured represent the process? If your samples are from a range that is low, % study variable is worse than it would be for the process. If samples can from too wide a range, %study variable seems better than it should be.
Be careful with taking full range of tolerance measuremesnts. Ideally you should have sample that represent the process being measured.
If not try using an historical standard deviation to get %Process.
If the samples do not represent the process that will be measured by the gage then the %Study variable cannot be applied to measuring the process. Basically %study variable is can I measure the samples in the study.
If you take samples to cover the tolerance width you can increase the variation of the samples to greater than the process.
For example. Take a process with a mean of 0 and standard deviation of 1. The worst possible scenarios for sample collection would be either all parts are 0, or you have 5 at -3, and 5 at +3.
The population has a sd of 1. In the scenario of 5 samples at -3, and five at +3, you will get a standard deviationof over 3.
A very important question to ask is do the samples they have measured represent the process? If your samples are from a range that is low, % study variable is worse than it would be for the process. If samples can from too wide a range, %study variable seems better than it should be.
L
Laura O
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
Allattar,
I really appreciate your response; it explains the use of the range very well.
Suppliers have modified parts to use more of the range of tolerance. Is there a % of the range that we should modify part to (say 75%) to insure we are using enough of the range to verify the gage? Historically we know that a certain process usually ranges +/-1 even though the tolerance is +/- 1.5.
Thank you
Allattar,
I really appreciate your response; it explains the use of the range very well.
Suppliers have modified parts to use more of the range of tolerance. Is there a % of the range that we should modify part to (say 75%) to insure we are using enough of the range to verify the gage? Historically we know that a certain process usually ranges +/-1 even though the tolerance is +/- 1.5.
Thank you
Last edited by a moderator:
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
you - and your suppliers - should never modify parts for an MSA. Use the parts you have available. IF you have historical data on the SD you can use that. The most critical reason for using the full range of the process variation is to look for interactions between the parts and the measurement system: for example the larger the value the more measurement error I get. Not all systems interact with the part.
you - and your suppliers - should never modify parts for an MSA. Use the parts you have available. IF you have historical data on the SD you can use that. The most critical reason for using the full range of the process variation is to look for interactions between the parts and the measurement system: for example the larger the value the more measurement error I get. Not all systems interact with the part.
Re: % of Contribution or % of Study Variation? Using Minitab - Gage R&R
A small point of clarification: this applies to variables studies (GR&R), but not attributes MSA, for which there are times when modifying parts is necessary.
A small point of clarification: this applies to variables studies (GR&R), but not attributes MSA, for which there are times when modifying parts is necessary.
Similar threads
