Closed ajnafa closed 2 years ago
@ajnafa It is probably not the most efficient approach, but the code was setup to use the Cholesky decomposition of the random effect correlation matrix (Cholesky decomp matrices classically called L
) in all cases. This (understandably) does not exist when there is a single random effect in a block --- I forgot to fill in 1s in that case.
I think a fix has just been pushed. Would you test now and let me know how it goes?
From what I can tell that seems to have fixed the issue and I'm no getting the error. Thanks!
Great, I'll close the issue for now. Feel free to reopen if it turns out not to be resolved. Separately, I'd also be interested for any other feedback or missing functionality. I cannot promise to implement, but I'd like a sense of what sorts of uses people have for this and how it can be made better. I'm currently working on it to fill a gap for a few of my own projects, but that only gives a narrow perspective.
I'm trying to estimate average marginal effects for a societal growth curve model (see Fairbrother 2014) and I've run into an error when specifying the
newdata
argument. The model is as followsI then try to estimate an AME for
time
based on the syntax in the vignettewhich fails with the following error
Since the time slope only varies across countries there is no
L_2
parameter in the model but I'm not sure I understand why its looking for it in the first place?