Seeley, JohnParr, Nicholas2020-09-242020-09-242020-09-24https://hdl.handle.net/1794/25631In meta-analyses of prevention programs, findings of primary research studies are pooled to estimate an overall program effect, an approach generally offering improved statistical power and precision over analyses at the individual study level. Across studies, however, programs are often implemented with considerable variation in implementation quality, program components, assessment approaches, and sample characteristics. Differences across these and other aspects of a program’s implementation can induce between-study variation in program effects. Excessive between-study variation can compromise the utility of a summary estimate of program effect, as derived in meta-analysis, because the estimate can be unrepresentative of the broad distribution of effects across implementations of the program. Importantly, variation produced by observable primary study characteristics is often explainable using variables that represent study methodology, program design, and sample attributes, and utilizing approaches such as subgroup analysis and meta-regression can provide insight into study-level factors that moderate the magnitude or direction of program effects. While widely used, these moderation analysis methods have recognized statistical and interpretive limitations, in particular when there is an interest in understanding the interrelation of multiple potential moderator variables and their combined influence on variation in program effects, as well as their co-occurrence in typical studies implementing a program. To address these limitations, this dissertation describes and demonstrates a multivariate approach to moderation analysis in aggregate-data meta-analysis, which employs finite mixture modeling as its underlying analytic framework. Results of the approach suggest it provides insight into the co-occurrence of potential moderators in a sample of studies implementing a prevention program, and into how such co-occurrence relates to program effectiveness.en-USAll Rights Reserved.implementation sciencemeta-analysismixture modelingmoderationpreventionsubgroup analysisElectronic Thesis or Dissertation