[R-meta] reliability and meta-analysis

[Viechtbauer, Wolfgang (SP)]

Dear Catia,

You should take the dependency of correlations computed based on the same sample into consideration. In other words, if you have cor(x1,z) and cor(x2,z) based on the same sample of subjects, then these two correlations are not independent. Note that rma.mv() with random effects is not sufficient to account for this. The covariance between the two correlations should go into the V matrix. See help(rcalc).

Best,
Wolfgang

>-----Original Message-----
>From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On
>Behalf Of Cátia Ferreira De Oliveira
>Sent: Wednesday, 23 June, 2021 17:19
>To: R meta
>Subject: [R-meta] reliability and meta-analysis
>
>Dear all,
>
>I hope you are well.
>Would one be able to conduct a meta-analysis where you want to see which
>measure of the construct leads to higher reliability? These would result in
>multiple correlations from the same participants which would be able to
>represent with the rma.mv function, but is there anything I may be missing?
>i.e. if I use different ways of representing a construct and then do a
>meta-regression where I check which method contributed for the higher
>reliability, do I need to consider anything else?
>
>Best wishes,
>
>Catia
>
>--
>Cátia Margarida Ferreira de Oliveira
>Psychology PhD Student
>Department of Psychology, Room B214
>University of York YO10 5DD
>pronouns: she, her