Closed jpiaskowski closed 1 year ago
There's a more general question here. Many of the "downstream" prediction/estimation/diagnostics/plotting packages can handle some subset of mixed-model packages. broom.mixed
is the only one that's specific to mixed models AFAIK. This set springs to mind: performance
, effects
, emmeans
, ggeffects
, DHARMa
, marginaleffects
, gtsummary
, stargazer
, ... what should our rubric/criteria be for including these packages in this task view?
I'm not entirely certain, to be honest. It depends on the overall purpose of this CTV, I suppose? My view is that this should help others implement mixed models and extract needed output from them. I view emmeans
, marginaleffects
, etc as convenience wrappers for extracting BLUEs from mixed model objects. While it's possible to do that without those packages, they do make it easier. I don't know what those other packages do (e.g. performance
), but it seems like if a package solves a needed issue and has methods specifically written for mixed model objects (e.g. merMod
), then their inclusion is warranted.
added.
marginaleffects (for prediction/estimation)