Subscribe Bookmark RSS Feed

Fitting generalized linear mixed models

philbirget

Community Trekker

Joined:

Oct 29, 2011

Hi everybody,

I'd like to fit a generalized linear mixed model onto my data. It seems that the Fit Model platform does not offer this option- is there another way of doing this in JMP?

Thanks for your help

Philip

1 ACCEPTED SOLUTION

Accepted Solutions
susan_walsh1

Staff

Joined:

Jul 22, 2014

Solution

JMP does not have the capability to fit generalized linear mixed models. The JMP development team is aware of user interest in this functionality.

10 REPLIES
deer_dog

Community Trekker

Joined:

Nov 19, 2011

I am using JMP 7, but it does offer generalized models under the 'fit model' platform.

Fit model, then under "Personality" (at upper right), you find different options including "Generalized linear model". It then gives you the option to specify the the distribution (binomial, poisson or exponential). One can also go directly to logistic regression (binomial or ordinal) as a "personality" within the 'fit model'. Is this what you are looking for, or are you trying to find something else such as a mixed model (GLMM)?

philbirget

Community Trekker

Joined:

Oct 29, 2011

I am actually looking for GLMM as I need to introduce random effects and potentially even nested effects.

deer_dog

Community Trekker

Joined:

Nov 19, 2011

Ok. I looked into this question *a lot* in the past, although my memory of what I found is a bit old. Here's a try.

You can do ANOVA with random and nested effects in JMP. Once you put the variable into the 'x' box, you highlight it and click on 'attributes' and "random" is an option. But that assumes your y variable has a somewhat reasonable normal distribution.

You can't do other GLMM (poison, binomial) with random effects. It looks like you can when you set up the model within 'fit model'. But, when you hit 'run model', it tells you that the generalized linear model does not take random effects. At least that is the case with JMP 7. But I looked into whether I could do this with a more recent version (JMP 9) and believe the answer is still no.

I turned to Stata and SAS for that analysis (survival with families as a random effect). I found Stata easier to use, but it would be helpful to know SAS or R these days. That was a few years ago, but it seems the same versions of JMP are still available.

deer_dog

Community Trekker

Joined:

Nov 19, 2011

p.s. you can designate a factor as 'random' and also include it as a nested effect - again in ANOV, not GLMM.

philbirget

Community Trekker

Joined:

Oct 29, 2011

Thanks a lot , it's a pity JMP can't do that (yet). I definitely can't fit GLMMs with my JMP 8.

All the best

P

deer_dog

Community Trekker

Joined:

Nov 19, 2011

well if someone using JMP 9 knows otherwise and it can fit a GLMM, let us know

jp1

Community Trekker

Joined:

May 19, 2013

Hi,

Any updates on this threat for JMP 10?  Looking to run a GLMM.

Thanks in advance.

deer_dog

Community Trekker

Joined:

Nov 19, 2011

Unfortunately, JMP 10 still does not seem to have this capability as far as I can tell. Frustrating and I do not understand why the software has not been modified to handle random effects. I will pretty much have to switch to another software (R) for this reason. When running a logistic regression, I often need to include random effects.

jpmontagne

Community Trekker

Joined:

Oct 7, 2014

I'm using JMP Pro v12, and it seems 5 years later, this is still not possible. Have any JMP users found a way to run GLMM, maybe with R? Otherwise, STATA or R can run GLMMs.