[R-meta] dependency and correlations for subgroups

[Cátia Ferreira De Oliveira]

Dear all,

I hope you are all well!
I am running a correlational meta-analysis with robumeta (I now know I
should have used a auto-regressive meta-analysis but this is what I
pre-registered last year before I knew better). When conducting data
extraction, I sometimes encountered situations where the authors provided
correlations for the overall groups (e.g. people with and without
language issues), but also sometimes for subgroups within this groups
depending on whether they used different tests to assess a certain
construct (e.g. by school). I know this runs into the problem of dependency
as they are using the same participants multiple times, but my question
pertains to the fact that variables such as age will differ for these
entries (i.e. the groups and subgroups will have different age means). Does
this pose a problem when we want to look at the influence of age or is it
fine given that they all come from the same study and it will just be a bit
misspecified?

I am sorry if this was a bit confusing!

Best wishes,

Catia

	[[alternative HTML version deleted]]