search

cran-task-views / MixedModels

CRAN Task View: Mixed, Multilevel, and Hierarchical Models in R
https://CRAN.R-project.org/view=MixedModels
27 stars 9 forks source link

new package suggestion: mmrm #21

Closed jpiaskowski closed 2 years ago

jpiaskowski commented 2 years ago

from @wviechtb :

New package just released: https://cran.r-project.org/package=mmrm

Essentially, it fits models of the form gls(outcome ~ time + other_fixed_effects, correlation = corSymm(form = ~ 1 | subjectid), weights=varIdent(form = ~ 1 | time)) (although one can also use other structures for the error var-cov matrix), so this is what they mean by "marginal linear model without random effects". In one sense, one could regard this as a GEE approach. Alternatively, we could put this under the Specialized models with a new entry Repeated-measures (or something like that).

jpiaskowski commented 2 years ago

Would it fit under "convenience wrappers"? It doesn't appear to be introducing new functionality, but it is providing a wrapper for particular circumstances.

wviechtb commented 2 years ago

As far as I can tell, it's not just a wrapper around gls() (the way I phrased it was a bit misleading), but I didn't dig into the code.

jpiaskowski commented 2 years ago

I saw it as a wrapper for TMB, which is great, because that package is hard to use. There are already many packages that handle repeated measures in this TV and other example where specialized covariance are used (e.g. "spatial"). It might make more sense to redefine that as "specialized covariance structure" or something like that for spatial and repeated measures examples.

bbolker commented 2 years ago

A little bit more history/context here: I worked a little bit with the lead author of this package (Daniel Sabanes Bove) to try to implement Satterthwaite df computations for glmmTMB. We got somewhere, but we had trouble with the combination of Satterthwaite dfs and "zero-residual-variance" settings (i.e., implementing "R-side" covariance structures, in SAS terminology), so he decided to write this package instead that implements a certain class of models in TMB.

jpiaskowski commented 2 years ago

I"m in favor of a repeated measures section. We will need to sweep through the entire TV to find all those packages.

wviechtb commented 2 years ago

I also think this could be useful, although it is of course a bit tricky since essentially any package for mixed-effects models can handle repeated measures. So we should focus on packages that specifically target this type of application (like mmrm).

jpiaskowski commented 2 years ago

okay, I'll create a branch and see how it goes.

jpiaskowski commented 2 years ago

going to work on this now.

jpiaskowski commented 2 years ago

This was added under a new subsection for "specialized models".