[R-meta] Model averaging woes

[Viechtbauer, Wolfgang (NP)]

Dear Maggie,

So you don't end up with two completely unanswered posts, I will at least respond. Unfortunately, I cannot really contribute anything to the discussion, as I am not familiar with the article by Cade. I just added it to my 'to read' pile, which is close to toppling over, so it might be a while before I get to this.

But a quick scan of the article seems to suggest that the author is not saying that one should not do model averaging at all, but that some kind of scaling of the predictors is needed, for example using 'partial standard deviations', which is not a term I am familiar with. Some googling led me to https://doi.org/10.1080/00031305.1994.10476059 which might be helpful, although I cannot tell you whether the same idea generalizes to the meta-regression context. I see that VIFs are involved, which can be computed with metafor::vif(), but again, it would be important to understand the logic first before just applying the equations in a novel context.

Best,
Wolfgang

> -----Original Message-----
> From: R-sig-meta-analysis <r-sig-meta-analysis-bounces using r-project.org> On Behalf
> Of Slein, Margaret via R-sig-meta-analysis
> Sent: Friday, March 21, 2025 00:22
> To: Viechtbauer, Wolfgang (NP) via R-sig-meta-analysis <r-sig-meta-analysis using r-
> project.org>
> Cc: Slein, Margaret <maslein using zoology.ubc.ca>
> Subject: [R-meta] Model averaging woes
>
> Hi all,
>
> I posted a less clear question on here but have solidified my conundrum:
>
> With an ecological meta-analysis, I would have assumed that model-averaging
> models within 2 AIC of each other was appropriate? However, in reviews, I
> received comments citing Cade 2015 where they argue that we should never be
> model averaging coefficients from ecological models. I’m kind of at a loss.
>
> Is model averaging in this kind of analysis totally flawed? What are other ways
> to do model inference? And how do we make comparisons of across predictor level
> variables with model averaged results?
>
> Any and all assistance would be greatly appreciated.
>
> Cheers,
> Maggie