1
$\begingroup$

I am conducting a Meta-Analysis using metafor in R. I calculate the standardized mean difference between patient groups (PG) and healthy control groups (HC) in the self-reported utilization of rumination.

Now, I would like to conduct a Meta-Regression with the mean-difference btw. the 2 groups of each study in the BDI-II (Beck Depression Inventory II) as a predictor.

Is this possible? Meta-regression is usually conducted for study-level variables. Practically, I can calculate a BDI-Mean Difference for each study. But I think that the SD or SE also had to be considered when making such type of calculations.

My dataset has, among others, the following variables: n.pg n.hc bdi.pg.m bdi.pg.sd bdi.hc.m bdi.hc.sd rum.pg.m rum.pg.sd rum.hc.m rum.hc.sd.

$\endgroup$

1 Answer 1

0
$\begingroup$

Your approach of including a second outcome as moderator is a bad idea. This approach fails to accommodate measurement error in the independent variable. If you want to test the association between the two outcomes, a bivariate meta-analysis model is the way to go. The challenge is to obtain the variance-covariance matrix for each study. Check this paper: https://onlinelibrary.wiley.com/doi/full/10.1002/sim.5679

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.