MV_META: A SAS® Macro for Multivariate Meta-Analysis- Meta-analysis of multiple outcomes and multiple treatments from a single study require more sophisticated models than the typical meta-analytic models. Three different approaches have been suggested to accommodate dependent effect sizes: a multivariate multi-level approach (Kalaian & Raudenbush, 1996), a robust variance estimation strategy (Hedges, Tipton, & Johnson, 2010), and the traditional univariate random effects approach (Hedges & Olkin, 1985). This paper presents a SAS macro that calculates multivariate meta-analysis confidence intervals, mean effect sizes, and estimated effect size variances for each outcome variable given a sample of effect sizes and sample sizes. This paper includes a demonstration of the macro, sample inputs and output, and an examination of the accuracy and precision of the three approaches based on a simulation study.
|
| Paper Number: | 210-2012 | | Start Date Time: | 4/23/2012 14:00 | | End Date Time: | 4/23/2012 15:30 | | Categories: | Posters |
- Add to MyPlan
|