covariate in GLMM

[#1]

If I want to include a covariate in my model, which I am running using the GLMM add-in because the data are not normally distributed and I want to use a Poisson distribution option - is the covariate included as a fixed or random effect? I know when using the standard "fit model" I do not specify that the covariate has a random attribute, which is why I wasn't sure how to include it in the GLMM model.

Second question: If after running an ANCOVA I want to graph my LSM + error bars (SE), I save as a table and then what do I do in graph builder? I imagine it would be inaccurate to just graph the data from the original table and not the LSMs because I understand that the LSMs account for the covariate effect and are therefore different than the regular means, and the error bars will probably be different, too?

Thank you!

[#2]

Hi,

It is generally better to post different questions in separate posts. Also, if you can include some non-sensitive example data as a .jmp file attachment that really helps people in the community to help you.

On your first question, I think this is hard to answer without knowing more about what you are trying to model. I think it would be unusual to have a covariate as a random effect but there might be a reason for doing this.

I assume that you have other effects that you are modelling as random, otherwise you shouldn't need to use the GLMM add-in. There is a "Generalized Linear Model" personality in Fit Model for Poisson and other distributions. There is also the Generalized Regression personality in Fit Model in JMP Pro.

I hope that helps,
Phil