topic Re: Bias correction factors in Discussions

Bias correction factors

Jendb75 — Fri, 09 Jun 2023 00:51:09 GMT

I am currently using an excel spreadsheet to calculate the results of an MSA. I am looking for software to spread up my process and I am currently trialing JMP. The calculations I am using in my spreadsheet are from the EMP III book by Donald Wheeler. My problem is that my spreadsheet repeatability and reproducibility results don't marry up with any of the JMP analysis options - Variance Components, EMP Gauge R&R results or Gauge R&R - despite getting the same average range, average, and control limits (on both the average range change and the range chart). I am wondering if I am using the wrong bias correction factors or the wrong formulas. I have attached my excel spreadsheet. Any help to understand the difference would be much appreciated. Thank you!

Re: Bias correction factors

ledi_trutna — Wed, 22 Jun 2022 13:34:51 GMT

I'm glad you are trying JMP. The EMP III book by Wheeler was used in developing some of the options in the MSA platform.

I think the difference in the results that you are seeing are from 2 things - JMP uses standard deviations instead of Rbars to estimate the variance components. Your spreadsheet is showing an Rdoublebar divided by d2 to get the within variation. And when using the bias correction factors, JMP uses the numbers to 10 decimal places, not the normal 3 that we often see posted in papers, and in your spreadsheet. So d2 to 10 decimal places will give a different answer than 1.128 for d2. You might want to check out our free course called Statistical Thinking for Industrial Problem Solving. We have a module on Quality Methods, and part of that is MSA. It will explain some of the calculations. https://www.jmp.com/en_us/online-statistics-course.html

Re: Bias correction factors

Donald46 — Wed, 29 Jun 2022 06:55:10 GMT

Hi Ledi,

I can't find any detailed information about the calculation method JMP is using for Gage R&R in this course. Is there any other resources that are available here?

Thank you!

Re: Bias correction factors

Mark_Bailey — Wed, 29 Jun 2022 14:32:39 GMT

You can search JMP Help for information. It includes the Quality and Process Methods Guide. See this section about the statistical details.

Re: Bias correction factors

Jendb75 — Thu, 07 Jul 2022 04:30:18 GMT

Hi Mark, I have looked through all the documentation and completed the MSA section of the free course. I also understand that there will be small variations in the results due to decimal places etc. However, what I can't work out is why the repeatability in JMP (0.604) is significantly different to both my manual calculations (0.281) and that of another software package I am trialing (0.277). I like the JMP software and would prefer to use it, however, this is not an option until I understand why the repeatability result is so different.

Re: Bias correction factors

Mark_Bailey — Thu, 07 Jul 2022 13:12:55 GMT

What are the factors in your study? What is the model that you are using?

Re: Bias correction factors

ledi_trutna — Thu, 07 Jul 2022 13:54:49 GMT

Attached is a journal I have called Repeatability journal. It walks you through calculating the variance components by hand for a simple study with just parts and within, and a more complicated one with instrument, operator and part.

Re: Bias correction factors

Jendb75 — Mon, 11 Jul 2022 00:38:22 GMT

Hi Mark, I have 3 operators (X, grouping) and 4 parts (Part, Sample ID). Each part is tested 3 times by each operator (Y, response). My data is attached to my original post. I have tried various models in JMP. The range dispersion chart type and Main model type give me a repeatability of 0.604. Using the crossed model type, the repeatability is 0.543.

Re: Bias correction factors

Mark_Bailey — Tue, 12 Jul 2022 13:04:11 GMT

The different results (0.604 versus 0.543) is that the former result is from a model without the interaction term and the latter result is from a model with the interaction term. You can't compare estimates from different models and expect agreement.