Heterogeneity
We estimated proportion of heterogeneity relative to sampling error
(I2; Higgins and Thompson 2002) and partitioned it
into between-study heterogeneity and within-study heterogeneity
(Nakagawa & Santos 2012). Current formulations of I2do not accommodate sampling-error covariances for multivariate
meta-analytic models. We, therefore, fitted simpler models with only
observation-level variances to estimate I2. While this
is not ideal, we note that meta-analytic effects of moderators
accounting for sampling-error covariances are robust to these simpler
models after adjustment with RVE (see Supporting Material).