[R-meta] robust variance estimation with few clusters

[James Pustejovsky]

Hi Filippo,

Your approach is to fit a univariate model and then cluster the standard
errors to allow for dependence across effects nested within papers. RVE is
based on using working models for the dependence structure, and so we might
call this approach an "independent effects" working model. Your approach
seems totally reasonable to me, especially if the higher-level structure is
difficult to estimate (because of few effects per cluster).

James

On Thu, Jul 25, 2024 at 8:08 AM Filippo Gambarota via R-sig-meta-analysis <
r-sig-meta-analysis using r-project.org> wrote:

> Hi,
> I'm approaching the robust variance estimation using clubsandwhich and
> following the metafor documentation, however I'm not sure about a point. I
> have a dataset with a multilevel structure (independent effects within a
> paper). However the number of clusters is low as well as the number of
> effects within clusters.
>
> My idea was to fit a standard two-level model ignoring the multilevel
> structure (author/id is the paper/effect structure):
>
> ```
> fit <- rma(yi, vi, data = data)
> robust(fit, cluster = author, clubSandwich = TRUE)
> ```
> However in the metafor documentation the robust function is used on a
> multilevel model. Do I have to fit a multilevel model (with the problem of
> having few clusters) and then the robust function or my approach is
> correct?
>
> thank you
>
> --
> *Filippo Gambarota, PhD*
> Postdoctoral Researcher - University of Padova
> Department of Developmental and Social Psychology
> Website: filippogambarota.xyz
> Research Groups: Colab <http://colab.psy.unipd.it/>   Psicostat
> <https://psicostat.dpss.psy.unipd.it/>
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-meta-analysis mailing list @ R-sig-meta-analysis using r-project.org
> To manage your subscription to this mailing list, go to:
> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>

	[[alternative HTML version deleted]]