A comprehensive multilevel model meta-analysis of self-concept interventions

![CDATA[The efficacy of self-concept interventions has previously been examined through traditional meta-analytic methods, and a host of moderators of intervention outcomes have been identified (O’Mara, Marsh, & Craven, 2004; Haney & Durlak, 1998; Hattie, 1992). However, traditional meta-analytic models have increasingly been criticized because they fail to account for the nested structure of effect sizes within studies, thereby violating statistical assumptions of independence. The multilevel model approach to meta-analysis, however, takes into account the hierarchical structure of meta-analytic data, thereby providing findings that are more statistically sound. Consequently, the present study applies the multilevel model technique to the analysis of the self-concept intervention literature. The overall mean effect size of .47 suggests a moderate impact of interventions on self-concept at posttest, and analyses showed that intervention effects are maintained at follow-up. Other moderators examined include the construct validity approach to the multidimensionality of self-concept; the use of random assignment to treatment and control groups; the control group type; the focus of intervention on self-concept; treatment characteristics; and the type of treatment administrator. Intra-study correlations and the variance explained by each moderator model are presented to emphasise the importance of using a multilevel model approach to meta-analytic research. It is concluded that multilevel models provide a more accurate understanding of the self-concept intervention literature than traditional meta-analytic models. Suggestions for future self-concept intervention design and evaluation are provided.]]