clarification regarding likelihood-ratio test in Fixed and Random Effects Models snippet

[davedgd]

@strengejacke: thank you so much for your awesome examples and packages.

I wanted to bring to your attention a potential question regarding your example here, namely in the section Check if a REWB- or simple RE-model suitable with the following ANOVA comparing fixed effects using REML with glmmTMB: anova(model_re, model_mundlak)

Rather than repeat all the details, perhaps you could take a look at the discussion I've posted here regarding the glmmTMB documentation: https://github.com/glmmTMB/glmmTMB/issues/686. The short version is that LRTs conducted with glmmTMB, unlike lme4::lmer, do not refit models with ML prior to producing the LRT output, which may be an issue with regard to the LRT.

I'm curious about your thoughts/interpretation.