How to do MSA when a single part also varies.

Typically, a MSA study would consist of say 10 parts measured 2 times by 3 people. What I am about to say is a generality in the absence of a lot of detail in your post. Since the parts are the same, the variation could be attributable to the Repeatability of each operator and the Reproducibility between the operators. This is called a crossed study since the parts are measured across the operators. If the parts also vary then you have a different situation much as you would have if the test you were measuring was a destructive one thus making repeated measurements impossible. In this case you would analyze as a nested study. The Reproducibility issue becomes moot but the part to part variation and the repeatability remain. The big challenge is to get parts that are similar enough in nature to allow the Repeatability to become valid should there be an issue.

This sounds pretty cool.
Since you mention that both the parts and the measuring are leading edge, is it possible to effectively split this into a 2-part study, meaning you could potentially use a previous generation of part that might be more stable or at least better known to pin down the measurement variation? From there, part 2 would be to use the now mor efamiliar measuring tool on the variable parts. At least that way you would have a means to check stability within the measuring. Seems this needs to be addressed before plunging into crossed/nested designs, etc.

I told you how to deal with it – think about it.

That’s not true Stan – those are your rules, which you fervently apply to everybody except your buddies.
The Forum rules say don’t “promote” a product or service. They are fine with responses that are applicable to a poster’s question and may lead to the parties doing business together; ie, Adam Bowden, Mike Carnell, Eric Maas  & others posting their company’s name, business email addresses and phone numbers. Forum has no problem with that…as evidenced by them not omitting them.
Let the Forum moderator do her job…and you do yours… of just criticising everybody for their stupidity.

BTW I gave the guy correct advice and the only thing I said to Mr. Star Wars is don’t complicate an easty problem.

Remi,With all due respect, your method buys him nothing. Declaring the
real value by an average does not make it the real value when you
just go and do single readings in an R&R. The original variation is
still there and you still are not sure where it is from.This is a problem that is dealt with daily in measuring electronics –
neither the measurement nor the measured device is “leading
edge”. Go use an oscilloscope to measure and vary the time period.
You will see that you go from something that is nothing but
variation with a short time period to something that is a rock solid
signal with a long time period. All that is going on there is
averaging. This guy hasn’t got a clue where his variation is coming from. It is
not a problem of running a measurement study, it is a problem of
how he will measure. The averaging of multiple readings will allow
him to reduce the overall variation so that he can distinguish
differences between parts. It is the answer to the measurement
problem in the short term.If he wants to validate the claim that the parts are varying, he will
need a better measurement device. The problem is more likely
variation across the parts which can be validated by presenting the
part to the measurement device exactly the same way repeatedly
and assuring the test is run the same way repeatedly.

Hacl,If the coffee you have does this for you, I’d like some of what you are
having.

HACL,  It sounds like you are putting together a classroom exercise!
So to Stan’s reference – for any one of the ten chutes, to measure repeatability, the one operator must be sure to place the ball in exactly the same spot across the 1-foot wide chute each time a measurement is made.  And to evaluate reproduceability, that same spot must be used from operator to operator as they measure the chute-time.  SOPs on HOW to measure.
Going back to the original question – if the measured value of an individual part changes from measurement to measurement, is the actual property of the part known to be changing with time?  Is it unstable? (measuring a block of ice on a 90dF day, 15min apart)  Or is it just that you are placing the balls at different points on the chute – so to speak?

What is the difference between Gage R&R and MSA? Does it assume annual calibrations?

Hai Petros,
depends on the person you talk to.

‘lazy’ person: they are the same
otherwise: Gage r&R is part of MSA: the study on repeatability and Reproducibility. Before that you should have checked: Validity; Resolution; Bias (also Linearityand Stability).
Remi

Wrong.
I, amoung others on here, have been through this dozens of times.
I told you what the solution looks like. Think about it.

I had a hard time finding it too through all the banter (though not as bad as some threads):
https://www.isixsigma.com/forum/showmessage.asp?messageID=153437
Can you disclose to us what this “part” is?  It’s rather intriguing.
An icicle?  A black hole?  Jar of wind?
Is the change over time linear, logrithmic, random?
I know – you fall into a catch 22 – how can you analyze the change if you can’t depend on your measurement to begin with.

CI Guy,
It is either in the way you are presenting the test item to the test and you need to be able to do it consistently or you need to take multiple measures and use the average.
“It’s easy” and you will not understand until you “think about it” instead of engaging in all these conversations about how hard it is.
If you don’t believe me, ask Mike Carnell.
If you want help from me, write [email protected].
If you keep on wanting to make this hard, I’ll keep telling you “it’s easy” because it is. I can solve this problem in 2 hours or less. You’ve wasted more time on here than that telling everyone how you are different and it’s hard for me – has anyone wever seen this?

BJ,
I have been thinking about this for a while and I don’t even remember the thickness on any of the metals except the gold (which was the one we got beat up for if we went through gold targets to fast).
No idea what measurement device we used. I have a connection to the guy that ran the lab for us on Linkedin. I can check with him if you really want to know.
Regards

?,
And that would be just your opinion. What is your point?
It has certainly made a difference in the impact of your post by using the term muda instead of waste. I am always impressed when someone speaks to me in a language I don’t speak.

Mike,Thanks for jumping in.Sorry for subjecting you to the dufus ?.I think CI Guy is invested in the fact he is different and this is an
incredibly hard problem. To bad, can’t learn when you are not
listening.

I know – I just thought you wouldn’t know.

I thought your analogy was special and especially well thought out for
the hour of the morning. I need some of whatever allowed you to do
that. If therapy is the answer please don’t point it out in public.

I was there too and I don’t remember either. Too many years and too many other companies and people that were
different with “leading edge” unsolvable problems.None of them were right by the way.

Bj,We are all very proud of you. Thank you so much for your special contribution. The next time you have another self esteem issue please be sure to stop back by with another rehtorical question.
.

I am not offended, just amused.

BJ,
I am not sure how you determined anyone was offended. Like Stan I am not offended just have no idea what point you thought you were making that would help the original poster understand that he had a possible solution and wasn’t willing to try it. If you want us to be impressed that you know the name of the measuring device – I don’t understand that thought process either.
So what is your point?
 

Stan I think the problem is your original answer was too simple. I wonder what the GR&R results are.

CI, let it go. As you can see through this 50 or whatever response chain you will not get answers here. You will get only “I’m so much better than you” comments that somehow make this little tribe feel better about themselves. They have their little pond with the big fish and the little fish who brown-nose them and it goes on and on. Trouble is the water in this pond is evaporating so they’re fighting for some form of “I’m still important”. Kind of a therapy group…”Hi my name is Stan (although not really) and I’m a has been who tries to act as if I still matter.”

I agree Les, my supplier assessment already told me what I need to know. It just confirms what someone told me about Florida.

This may be a bit too obvious and too simplistic an answer and if so, I apologize.
If you are testing new parts with a new measurement device, is it possible to test the device in a fixed, controlled environment with non-varying “dummy” parts first? Use a “proxy” for the real part that you want to measure but one that does not vary from observation to observation. Do a full gage R&R study in that environment and determine how reliable your measurement device and measurement procedure is. Once you are satisfied, then turn to your real parts. You should then be able to accuratley subtract the measurement system variation from the part variation.
Sorry if that is too simplistic and maybe there isn’t a suitable non-varying proxy that you could use.
Just my thoughts on the topic.
 

Assuming you are not willing to take Stan’s advise which is sound, another poster has already suggested an alternative approach.  You can utilize a “nested” approach.  To read up on this, visit http://www.keithbower.com .  He gives you advise on how to set it up and how to identify if the assumption of homgeniety (sp?) is violated.
Having said that, it is a more complicated approach and the simple solutions Stan has suggested will be more effective.

Raja,
Good reference article, but I don’t think this issue is “within-piece” variation. I’d be surprised if CI guy could sub-divide the unit under test and take this approach…but only he can answer that.
CI Guy,
I had a similar issue with trying to characterize a torque analyzer. The “parts” in the study were torque wrenches. We had a hard time discerning whether the wrench was delivering a consistent amount of torque or whether the operator’s technique varied, or whether the gage varied. We agreed on a technique of 3 repeat measures by the operator, and used the average (just as Stan has suggested here). This is how we monitor the wrenches in production. (Qualify the tool with average of 3 readings…this is how you must use it!)
Keep in mind that averaging can be a problem if your data is non-random. I have seen this with SEM equipment for measuring critical dimensions on wafers. The data is highly auto-correlated due to the build up of surface charge from measurement to measurement. Before using this approach..make sure the measurements are not autocorrelated. (We could explain this observed variation, but I am still not sure how your part varies. I also know that if I run a GRR on my bathroom scale, my beer drinking will impact the weight measurements. I also step on the scale without any help, so I am the operator and the part!   There’s that weird coffee again.)
If I were you, I would run a quick multi-vari study.
Days: 1 through 5
Operator (within-day): A, B, C
Part (Within operator, Day) 1,2,3,4,5
Repeat (1,2,3,4,5)
The error term in your nested ANOVA will represent the gage error and your suspected “noisy” part variation. I am not a salesman, but for this analysis I would use JMP. You will see if your data is autocorrelated, you can plot the data on a control chart and assess the stability of the measurement system, etc. Why is everyone so biased towards R&R? (Get it…that was an MSA joke!)
MSA = Linearity, Stability, Bias, Repeatability, Reproducibility

CI Guy,Here’s hoping you learn who actually gives good advice some day.Come back and tell us what you learn when you are done.I could tell you about your parts and measurement in 2 hours or less.
Anyone who has taken appropriate training on MSA knows how to
deal with this.