nominal logistic and generalized linear models do not take random effects

l_yampolsky — Fri, 09 Jun 2023 00:27:21 GMT

Dear JMP Community,

Sorry if this has already been discussed, I think it has been, but I don't seem to find the relevant thread.

I am attempting to run either nominal logistic or generalized linear model with binomial distribution (in this particular case if I am not mistaken these are pretty much the same thing) with a categorical response variable and a nested design, in which I have the outer factor I am most interested in and a nested inner factor which is basically part of random variation. Of course I want to test the outer factor against the inner one, so I specify the nested inner factor as random. But I cannot, because these models do not take random effects.

Is there a fundamental reason why random effects cannot be evaluated in these models? Perhaps this option exists in newer versions of JMP? (I'm still on JMP 10). Or in JMP Pro?

If not, what are my options? I do not want to pool / ignore the inner factor because there is for sure a lot of variance among its levels and I am inflating my power.

Maybe I should do some sort of resampling, randomly choosing 1 replicate from each nested factor category?

Many thanks for any suggestions!

Re: nominal logistic and generalized linear models do not take random effects

l_yampolsky — Mon, 11 Jan 2021 13:54:49 GMT

I found this:

which is exactly what I was looking for, because it does include random effects.

Can't run it on version 10 though..

topic Re: nominal logistic and generalized linear models do not take random effects in Discussions

nominal logistic and generalized linear models do not take random effects

Re: nominal logistic and generalized linear models do not take random effects