Skip to content

[R-meta] question on multilevel model results

4 messages · Michael Dewey, Gabriel Cotlier

#
Hello all,

I would like to ask why, conversely, rma() objects for rma.mv() objects
(with and without moderators ) in the summary report do not provide values
of heterogeneity measures besides Cohen's *Q* such as "tau/hat/ square" or
"tau/square" or H^2?

Is this maybe because this type of model (rma.mv()) does not allow the
estimation of these measures?

Alternately, could they be calculated separately for rma.mv() objects?
If so, is there a code for doing this ?

Thanks a lot.
Kind regards,
Gabriel
#
Dear Gabriel

Does 
http://www.metafor-project.org/doku.php/tips:i2_multilevel_multivariate 
help?

Michael
On 27/07/2023 08:18, Gabriel Cotlier via R-sig-meta-analysis wrote:

  
    
  
#
Dear Michael,
Yes, Thank you!!
It is indeed very useful for I^2 in multilevel models.
But I was also wondering how to obtain Tau and Tau-squared in multilevel
models?
And if such a thing is possible?
Kind regards,
Gabriel




On Thu, Jul 27, 2023 at 1:22?PM Michael Dewey <lists at dewey.myzen.co.uk>
wrote:

  
  
#
Hello Michael,

I assume that tau^2  in multilevel models is divided into tow:

1.  tau^2 _1 for level 3 that is between-cluster variance equal to  between
study heterogeneity
2,  tau^2_2 for level 2 that is within study variance or within study
heterogeneity.

Is this correct?

Thanks a lot.
Kind regards,
Gabriel
On Thu, Jul 27, 2023 at 3:31?PM Gabriel Cotlier <gabiklm01 at gmail.com> wrote: