MSA ANOVA GRR

[#1]

Hello Community,

It has been a while since I last did a GR&R in JMP. Back then, I remember I had the option of selecting if the analysis should be based on ANOVA or Xbar. I don't find this option anymore and I cannot get the same results as my peers using Minitab for their ANOVA GRR.

This is sparse info, I admit, but can someone tell me how I can ensure I do my GRR based on ANOVA?

Thanks in advance

Lars

[#2]

What version of JMP?  In the current version, if you go to Analyze>Quality and Process>Measurement System Analysis there is an option to use EMP (Wheeler's approach) or Gage R&R (a more quantitative approach).

[#3]

Hi Statman,

I am on JMP 16. I know about the options you refer me to and I understand the EMP is a more modern MSA analysis and what I understand preferred. Is there one of those that can be compared to Minitab's ANOVA GR&R? I seem not possible to achieve the same results as my colleagues on Minitab. Thanks.

[#4 · Accepted solution]

You can also still use the Variability Chart for Gage R&R. Please see the documentation here.

Your primarily goal seems to be obtaining the same result as MINITAB for compatibility. So pay attention to the options in the Variability Chart for how the variance components are estimated. You might have to change the procedure if the default method is not what you need.

[#5]

Hi Mark,
I ran the analysis using Variability Gauge Chart. For Analysis Setting I chose REML and got pretty darn close to Minitab's results and will conclude the results are the same.
On the question to ANOVA. Here it's my ignorance that as long we have the operator-part interaction we choose to call it an ANOVA.
BTW, the study is related to continuous data collected from a torque gauge, setup as a traditional 3 operator, 3 repetitions on 10 parts.

[#6]

Dear Community,

even though this chat is a few years old, I would like to reopen the topic. I am facing the same problem described in this chat. I now found out that minitab uses the EMS-Model to calculate the variance components for the GR&R. However, JMP automatically selects REML for the model. In the settings for the MSA I can only choose the option "Choose best analysis (EMS or REML)" and then JMP always choses REML. When do fit model and standard least squares, I get the same result for the variance components as in minitab (or "by hand" in Excel). I somehow have to get JMP to use the model I get from the standard least squares fitting for the GR&R.
Does anybody know how I can force JMP to use EMS in the MSA?

Thanks in advance,
Annika

[#7]

Can you attach your data table? Is the purpose of this query to get JMP to match Minitab or to do the most appropriate analysis?

[#8]

Hi statman, 

I attached mock data that describe the problem: I am looking for the same result as in Minitab, probably, the REML model that JMP uses is a better fit for the model, however due to company guidelines I need to stick to the EMS model for GR&R. Minitab uses the EMS model. 

In my example (GR&R_TestData.jmp) I have a "classic" crossed GR&R design 5 parts, 3 operators, 2 trials. Minitab does the ANOVA using the EMS model and then calculates the variance components using standard GR&R formulas:

In JMP I get the same ANOVA table, when I use the EMS model in fit model:

However, when I use the  Variability Chart for Gage R&R in JMP, JMP automatically uses the REML model, resulting in (slightly) different parameter estimates/variance components.

If jmp would use the MS values from the EMS model for the GR&R, I would get the same results as in Minitab. Thus, I need to force jmp, to use EMS instead of REML. However, in the analysis settings in the Variability Chart for Gage R&R in JMP I cannot choose EMS. JMP choses automatically. 

I hope this helps to understand my problem.

Thank you very much in advance and best wishes!