How variation is calculated for REML and Bayesian models?

[#1]

I'm performing GRR study for destructive test. Two instruments, each measures unique set of parts, each part is measured 4 times.
How does program find repeatability and reproducibility for the set-up? Mean of the set measured by 1st instrument is 1.493um, by 2nd - 1.484um. tolerance is 0.2um.However, it shows 13% variation in reproducibility according to REML, where did this number come from?

[#2]

Is it possible you can provide the complete data table rather than one repeated measure?

https://www.jmp.com/support/help/en/18.1/?os=mac&source=application#page/jmp/variance-components.sht...

I'm sure you realize when using destructive (or altering) measurement systems, you cannot mathematically separate product variation from the repeatability precision of the measurement system.  There are strategies to bias the repeated measures to the measurement error by selecting samples for the study that are likely to have little variation in them, but specific assignment of the measurement precision repeatability is impossible.

[#3]

REML estimates variance components; 13% reproducibility reflects instrument differences.

View more...

Interesting thread! I’m still learning the deeper stats behind Bayesian models, but this reminds me of assessing risk at the poker table—probabilities always come into play. Just like https://casinoveri.it/ helps evaluate trustworthy platforms based on real criteria, understanding model variation helps you make better decisions in research (and gambling!).