POV (partitions of variation) vs COV (coefficients of variation)

Question

Hi,

I have someone asking me if we do Partitions of Variation natively in JMP, or have something similar.  I can't find much of anything online.  Apparently, it's used in semiconductor "when you are trying to solve a problem, one of the steps that is important is to partition out the sources of variation."  POV adjusts for handling continuous and attribute data along with balanced and unbalanced data.  "It also breaks out the within and between variation, including common / noise."

Does anyone know if this is included in JMP and how it compares to the standard Varience Components in Variability/Attribute Gauge platform?

HadleyMyers · Answer

Hi,

Thanks for the responses.  I think I've got enough to take back with me for now.   Thanks again.

Best regards,

Hadley

Chris_Kirchberg · Answer

Another variant to this is Principal Variance Component Analysis (PVCA) where you create Principal Components (PCA) from the responses, then use Variance Components like Jason has shown above but model against the Principal Components instead.

Here is a link from Russ W.'s post from about 11 years ago regarding this analysis PVCA

Also a link from NIEHS that might helps as well NIEHS PVCA

Jason_Wiggins · Answer

Hey Hadley,

Is it possible they are talking about variance components analysis? If so fitting a model with random effects will generate variance components estimates

Within variation is the residual in the Variance Components Estimates table. Between is available for an interaction model. Total gives total variance. Before I was a JMP user this is how I solved MSA problems. It is handy for any circumstance where there are random effects as well. Could this be what they are asking for?

J

sorry Mark, we must have been replying at the same time