Explaining heterogeneity in meta-analysis: a comparison of methods.