Analyses will be combined using hierarchical random effects. This analytic approach will
allow both an estimate of the overall (population) effect, as well as an estimate of the
variance of the effect across studies. These estimations are preferable to the use of an
arbitrary variance cutoff value or statistical tests for heterogeneity, such as Q statistics or
I2 scores. The decision of whether to partially pool a set of studies in a random (or mixed)
effects meta-analysis depends not on how heterogeneous their outcomes are, but rather on
whether they can be considered exchangeable studies from a population of studies of the
same phenomenon. This decision should be based on the design and quality of the
studies, independently of the relative effect sizes of the studies. Newer approaches to
random effects meta-analysis allow for robust (e.g., nonparametric) estimates of variation
that do not rely on the assumption of normal random effects. This permits us to account
for “outlier” studies in the meta-analytic model without discarding them unnecessarily