search

strengejacke / ggeffects

Estimated Marginal Means and Marginal Effects from Regression Models for ggplot2
https://strengejacke.github.io/ggeffects
Other
560 stars 36 forks source link

Request: could you make ggeffects take objects from metafor (class rma.mv rma)? #30

Open itchyshin opened 6 years ago

itchyshin commented 6 years ago

Dear SJ

I was wondering whether you could get your wonderful package to work with rma.mv and rma objects? I just recently came across (yesterday) your sjstats and this package so wondering how difficult it would be to do this for the metafor package (a package for meta-analysis). In a meta-analysis, we often have a lot of categories so it would be good to know marginal effects but currently, this is not possible (at least easily) with metafor. Anyway, many thanks for these cool packages!

Thanks!

Shinichi

strengejacke commented 6 years ago

I haven't worked much with meta-analysis and prediction, so I think I must get a bit more into this topic to know what I'm exactly doing.

itchyshin commented 6 years ago

@strengejacke

Meta-analytic models are versions of mixed models. However, metafor's rm.mv may be written in a very different way, say, from lme4's lmer. I wrote a paper explaining this similarity. Please look at here.

https://www.researchgate.net/profile/Shinichi_Nakagawa2/publication/233341316_Methodological_issues_and_advances_in_biological_meta-analysis/links/00b495157aec0585c0000000.pdf

I hope this gives you an overview of normal meta-analytic models as well as multi-level (hierarchical) meta-analytic models.

Any questions or anything I could help you with, let me know

Thanks!

Shinichi

niekverw commented 3 years ago

I was also wondering how to use ggeffects with meta-analyses. I have very small datasets, that combined in meta-analyses, show significant interactions between covariates but not on their own; I would like to plot the predicted effects when combining each of these smaller datasets (I can't append the datasets due to privacy etc).

Alternatively, would it be possible to combine multiple ggeffects objects using meta-analysis, and then plot that?

jukicivan commented 3 years ago

I just wanted to add that this would indeed be a really nice add-on to this wonderful package. I've heard this from many people as well.

I think it would also be useful to let ggpredict work with coef_test from clubSandwich package after meta-analysis, and especially the multivariate meta-analysis model has been fitted. To my knowledge there is nothing out there like predict for linear models, or predict.rma.mv for multivariate meta-analyses, so we're limited with presenting graphs that should contain robust variances instead.

strengejacke commented 3 years ago

If anyone could provide a reproducible example, that would be helpful to work on this issue.