Skip to content

[R-meta] Determining if an rma.mv model is singular

3 messages · Stefanou Revesz, Michael Dewey

#
Dear Wolfgang and list members,

Is there any function or code for rma.mv models that would work like
lme4::isSingular()?

I ask this because, given a fixed number of moderators, I sometimes
want to empirically explore how much room for adding random effects in
rma.mv I have before my model becomes singular but I don't know how to
do that.

Thank you so much,
Stefanou
#
Dear Stefanou

Would it not be simplest just to fit the models and see what happens? 
You are not going to break anything.

Michael
On 22/03/2022 00:50, Stefanou Revesz wrote:

  
    
  
#
Michael,

Given the history behind the "sneaky" nature of singular models and
how easily people fall victim to it
(https://arxiv.org/pdf/1506.04967.pdf), I would think it may not be
advisable to simply fit the models and see what happens, indeed
nothing noticeable may happen at times
(https://stats.stackexchange.com/a/449122/140365) and that's probably
why lme4::isSingular() exists.

Stefanou
On Tue, Mar 22, 2022 at 8:25 AM Michael Dewey <lists at dewey.myzen.co.uk> wrote: