Search the Elsmar Cove!
**Search ALL of** with DuckDuckGo including content not in the forum - Search results with No ads.

Samples for GR&R - Can we get the 10 samples from multiple models


Registered Visitor
In GR&R study, we normally get 10 samples of the same model of part. Can we get the 10 samples from multiple models (in same product family) of parts with the same measurement system ? For example, my product family includes a bar shape part of different width. Say, model A, model B, model C, model D. Can i take 3 parts from model A, 3 from model B, 2 from model C, and 2 from model D together as the 10 samples for my GR&R study ?

b.rgds - dennis


Starting to get Involved
The selection of parts affects any characterisation of the relationship of the part variation to the measurement variation therefore you will need a faithfull representation of the part variation if this relationship is important for your needs.

Selecting parts from different families for use in a study such as this is fine if you are happy that the characteristics being measured are the same, including the geometry and size and that the manufacturing processes are the same. If the other part families are made on the same machines with the same process then the estimate of part variation will be faithfull for that process including whether they were made on different days and subjected to different environmental conditions.

You hint at using parts of different widths, again, is it reasonable to assume that the part and measurement variation should not change due to this change in width? If yes then you wont be reasonably estimating part variation and your measurement error estimate will reflect parts from different processes.

Hope this helps

John Predmore

Involved In Discussions
The GR&R compares the variation between appraisers to the variation among the parts. If you mix apples and oranges, your GR&R will quantify people's ability to distinguish apples from oranges, which may not be what you are after. If you plan to use your gauge to distinguish good apples from bad apples, all your samples should be in the range of good-to-bad apples. If you want to differentiate between good apples and better apples and not-as-good apples, it would be best to use this as your range of sample parts.


Registered Visitor
Yes. The family of parts are all made from the same production line under same process, eg. grinding. The only difference could be the grinding wheel size and the adjustment of the cutting depth. The main purpose of doing in this way is to try to cover more than one model in the study and the GR&R is assessed in term of %GRR instead of %P/T.
Actually I think this is similar to using gauge blocks of different length to do GR&R study. Some people do mentioned this method as I read from book / the web. Is it really not recommended ?

Bev D

Heretical Statistician
Staff member
Super Moderator
You can do anything you want, but “just because you can doesn’t mean you should”.
I’d start by understanding what you are trying to understand about your measurement system. Remember that the goal of any study is not ‘pass’ but to gain insight and understanding that allows you to make quality parts. The study you propose won’t really inform you of much at all.
Top Bottom