R&R Study MSA 4th Edition x JMP

[#1]

When I take the R&R data in the MSA 4th edition manual and put it in the JMP to redo the studies, the results don't match.

Has anyone done this validation, could they send me a video or images, I believe I'm doing something wrong.

Thanks

[#2]

First, welcome to the community.  It would be really helpful if you let us know which version of JMP you are using and what OS.  In addition, if you could post the JMP data table it would helpful.  There are multiple approaches to doing measurement systems analysis.  I am partial to the Wheeler & Lyday methods for analyzing components of the measurement system (e.g., precision repeatability, precision reproducibility, discrimination, stability).  You may certainly get some varying results depending on the method.

[#3]

Thanks for the reception. In the attached worksheet, on the R&R tab is the example of the 4th Edition MSA Manual, on the Worksheet1 tab is the example generated by the JMP. In theory should they have the same results, or am I doing something wrong?

[#4]

I have some comments in general:

1. Never fall in love with the statistics.  They are ALL estimates.

2. In the case you are looking at, I believe you have done one of the methods JMP will perform (EMP method) correctly.

3. The precision repeatability component shows evidence of inconsistency (out-of-control).  When this occurs, you have less confidence in any quantitative estimates.

4. I did not do a detailed evaluation of the R&R tab in excel, but if this analysis was done as a nested study and the analysis in JMP was crossed then nested, you can get different values for the components.  Additionally, JMP was probably using Bayesian estimates as the REML components were negative.  It looks like the R&R is using EMS method?

5. In general, it looks like the largest component is parts followed by operator.  I believe the results are similar.

I have attached the JMP file with scripts for analysis.  It is best to show the data tables in JMP  (vs. some other software) when working with the community.

[#5]

There are below basically three methods existed to perform Gage R&R. The results obtained from them for the same data set will be significantly different ?

Another doubt, Which method does JMP use ?

Is there any way to repeat the JMP results manually (using excel for example) ?

[#6]

I'm not sure what you mean by "below basically three methods existed to perform Gage R&R".  There are a number of ways to estimate variance components, and there are multiple "issues" to be evaluated with respect to measurement systems.  First a list of issues:

1. Discrimination (aka effective resolution).

2. Stability

3. Precision repeatability

4. precision reproducibility

5. Bias

6. Linearity

7. Accuracy

How you estimate each "issue" can be situation dependent.

For example, if you have a measurement device that lacks effective resolution, then it tends to round the data (too much) and therefore may give incorrect estimates.  If the measurement system is unstable or inconsistent, then the conclusions you draw from the study may only be right for that inference space.

As I have already stated there are multiple ways to estimate the components meant to be estimated by the Gage R&R. For example:

R-bar/d2 using control chart method

Expected mean square estimates (EMS)

Restricted maximum likelihood (REML)

Bayesian

Each of these will likely give different values for each estimate.

You state "results obtained from them for the same data set will be significantly different".  Please define significantly different.  I don't see the results as significantly different.  You would have the same rank ordering of components whether using JMP or the excel spreadsheet you shared.

I don't understand the question "Is there any way to repeat the results manually".  Pardon my confusion, WHY?  Why do you want to get the same results as your excel program?

You might enjoy reading Wheeler & Lyday "Evaluating the Measurement Process" or see here:

https://community.jmp.com/t5/Mastering-JMP-Videos-and-Files/Using-the-EMP-Method-for-Measurement-Sys...

[#7]

Thank you for your answers. I’m a starter on this topic, so, I can’t clarify how significantly would be the differences between the methods.

The point to define all this question is the following: I’ve got 3x different analysis from the same Data set (One Excel using Average and Range, JMP and another excel using ANOVA). From both I’ve got different answers, as you can see partially:

	Average and Range	JMP	ANOVA
VE =	0,2019	0,0382	0,0460
VP =	1,3905	0,8847	1,0887
VT =	1,4238	1,2544	1,1347
Gage R&R	0,3058	0,0649	0,0460

The reason I’m trying to reproduce the results on a excel program is to apply the theory from a MSA manual... The JMP will be the verifier for my evaluation.

Best Regards and thank you again.