hikea

Thank you very much for a quick reply! Regarding raw_rt, I did convert lexdec RTs back to milliseconds, so all my examples are on raw RTs. The entire question discussed...

I did the conversion like so in R (don't know how to do it in Julia so far) lexdec$rt_raw

Thank you! It looks like I get a proper fit for the Gamma model with Identitylink when I don't use fast=true ```Julia julia> lexdec_glmm_gamma = fit!(glmm(@formula(rt_raw ~ 1 + Class*NativeLanguage...

Not a problem. I've noticed some weird behaviour when using `fast=true`. Is it similar to `calc.derivs = FALSE` in lme4? When I use that, I can fit any model, fast,...

Unfortunately, I don't have anything to share at the moment, but I will when I encounter something again. Thank you for all the explanations. I hope this post is useful...

To address my third question, using my data (20k observations, 3x3x2 design), I dropped a by-word random slope (the maximal model for my data requires one) and used `fast=true`. This...

@bbolker thank you! This is very helpful. I've never done any implementations, but I'll take a look.

Looking a bit more at this (not necessarily in the right direction), I've noticed that correlations of random effects do not match. ```Julia julia> lexdec_model_rt_raw = fit!(lmm(@formula(rt_raw ~ 1 +...

I am glad that I am being of help here as well (although, I thought it was the opposite). I know that these are nonsensical models with wrong parameters, but...

Thank you @dmbates I've just noticed that Inverse Gaussian implementation was merged and it works in MixedModels (after using Pgk.checkout("GLM") and Pkg.chekout("MixedModels") ). However, I get the same error you...