Narrative reviews cannot systematically account for moderators—that is,
participant, treatment, or study design characteristics that influence the
variables under study. Meta-analysis uses moderator analysis to assess
such influences on variations in effect size. If moderator analyses indicate
that certain characteristics do not have an influence on the overall effect,
then these findings suggest that results are robust and may be transferable
to other settings. If the studies are inconsistent between settings, the
sources of variation can be examined (Glasziou et al., 2001). Recall that
Wilson, Lipsey, and Soydan (2003) found mainstream programs for
juveniles to be just as effective for minority and majority youth. In the
home safety education meta-analysis described above, investigators
found that these interventions were no less effective for families whose
children were at greater risk of injury (Kendrick et al., 2007).