What are the core packages
Closed this issue · 6 comments
Starting a list here so we can keep track of this. Core packages have to be on CRAN (which sadly, leaves out broom.mixed)
lme4
nlme
brms
MCMCglmm
multilevelmod
geepack
lavaan
FYI: I found out the hard way that every time a core package is mentioned in the task view, it has to say include the tag priority = "core" to be processed correctly by ctv::ctv2html(). I tried to fixed the instances of this in the TV, but I may have missed some.
Note that broom.mixed is on CRAN ... https://cran.r-project.org/web/packages/broom.mixed/index.html
I agree with everything on the list above. I might include performance and DHARMa (for model checking, although arguably not since they're not MM-specific); lmerTest (for LMM p-values); glmmTMB (broader range of response distributions, faster in many cases) blme, rstanarm (Bayes), ordinal, robustlmm (the most important of the 'special types', for my work ...)
Maybe the best thing would be to decide approximately how many core packages we want in total? (20?)
Good news about broom.mixed! I missed that. All those ideas sound great. There's a function in the ctv package to install all "core" packages, so keeping it to 15-20 sounds reasonable.
New list:
lme4
nlme
brms
MCMCglmm
multilevelmod
geepack
lavaan
broom.mixed
DHARMa
lmerTest
glmmTMB
blme
rstanarm
ordinal
robustlmm
The only one of these I'm not familiar with is multilevelmod (I've installed it and see what it is, I think I see why it would be included -- parallel to broom.mixed). I think I can see the argument/case for all the rest. (nlme might not count against the total as it's present in the default installation.)
yeah, currently {multilevelmod} is limited in scope, but I have high hopes that it will eventually support more functionality since it's backed by the tidymodels team.
I'm personally not in favor of adding emmeans as a core package because it doesn't do mixed modelling - it doesn't even provide BLUPs (while {marginaleffects} does). However, I don't feel that strongly about this.
I also wouldn't add emmeans as core -- its functionality is very general and not at all focused on mixed models / specialized to take advantage of what mixed models can do.