Yes, we know... We read YOUR guidelines <3 XD

Show Post

For the record I said that the (1|g) term is redundant in the *second* term, i.e. (1 | g) + cs(0 + f | g, hom = TRUE) ... (1|g/f) is identical to (1|g) + (1|g:f) (or (1|g) + (1|f:g)), and the (1|g) is *not* redundant ...

Show Post

I hope I am not getting lost here; if I am, please tell me. To the best of my understanding, the formulation `(1 | g) + (1 | g:f)` is not redundant, whereas `(1 | g) + (1 | g/f)` is redundant. So, no, we did not discuss redundancy.

Show Post

Correct—there are no heterogeneous correlations. I suppose every simulation study will always be at least one simulation short XD

Show Post

I personally agree in using by default maximum slope models, and only if they fail use CRIs. This is exactly our suggestion. I think that what you really lose is the opportunity to study and make inferences on the random effects.

Show Post