Chapter 24 of Borenstein et al (2009) addresses meta-analysis of 'multiple outcomes or time-points within a study". As far as I can tell, this still assumes that one variable is common across the effect sizes. I.e., IV on DV1 and IV on DV2. This contrasts with some situations where you might want to average IV1-DV1 with IV2-DV2 or IV1-DV1, IV1-DV2, IV2-DV1, IV2-DV2.
O'Mara and Marsh discuss existing options including (a) choosing just one effect size; (b) averaging effect sizes, (c) running separate meta-analyses. They conclude that
In summary, there are no truly satisfactory ways of dealing with multiple outcomes when the researcher is explicitly interested in comparing and contrasting different outcomes in the traditional meta-analytic models. In response to this, researchers are now exploring multilevel modelling approaches to meta-analysis.
References
O'Mara, A.J. & Marsh, H.W. (2008). Incorporating within-study correlations in multivariate meta-analysis: Multilevel versus traditional models. SELF Research Centre, University of Oxford: Oxford, UK (in review), , .DOC
Chapter 24 of Borenstein et al (2009) addresses meta-analysis of 'multiple outcomes or time-points within a study". As far as I can tell, this still assumes that one variable is common across the effect sizes. I.e., IV on DV1 and IV on DV2. This contrasts with some situations where you might want to average IV1-DV1 with IV2-DV2 or IV1-DV1, IV1-DV2, IV2-DV1, IV2-DV2.
I've implemented many of the functions here: https://github.com/jeromyanglim/meta-sem/blob/master/multiple-outcomes/multiple-outcomes.md
O'Mara and Marsh discuss existing options including (a) choosing just one effect size; (b) averaging effect sizes, (c) running separate meta-analyses. They conclude that
References