Sorted! the package agricolae was causing some incompatibility with the mgcv package.

Hello Catarina,

as I see you are also struggling with GAMs. May I ask YOU a question. I saw that you used the s(Site, State, bs="re") form. I have a similar situation in which I model an individual time trend, where individuals are nested in counties. Hence, I modeled a s(user, bs="re") and a s(county, bs="re"). I never thought to combine them. Have you any reference where dealing with such complex nestings is discussed? Or why did you combine them?

All the best and thanks in advance.

Holger

I am using it since I have a nested random effect. It makes more sense to do sites within States to me or by state.

:) Here the reference:https://people.bath.ac.uk/man54/SAMBa/ITTs/ITT2/EDF/WoodGAM.pdf

Hi Catarina,

when I try to nest the the RE, I get an error that R is not able to allocate the vector of 30 GB...Any idea what that means?

Thanks for the ref.

All the best,

Holger

Hi, difficult to know without looking at your code or knowing what data you are using.

I left my question because mgcv was having incompatibility with another package (see above), both had the same function name. That could be one of the problems? But I am guessing here. Maybe try to run mgcv::gam(x~y...) and see if the results persist. Another solution would be to ask a question on stack or crossvalidated.

:)