Meta-analysis is a favorite tool to summarize research findings in the social, behavioral, educational, management and medical sciences. Conventional meta-analytic techniques, which are known as the univariate meta-analysis here, assume independent on the effect sizes. When the effect sizes are not independent, a multivariate or three-level meta-analysis may be used to model the dependence. A multivariate meta-analysis is used when the effect sizes are non-exchangeable or measuring different constructs. A three-level meta-analysis is applied when the effect sizes are exchangeable. In practice, studies may include a mix of independent, non-exchangeable, and exchangeable effect sizes. Researchers usually have to choose among the univariate, multivariate, and three-level meta-analyses. This study shows how these effect sizes can be combined and analyzed in a single framework. The main advantage is that researchers may compare the average effects and heterogeneity across these designs. Examples will be used to illustrate the procedures. Future studies of the proposed approach will be discussed.