Invited Symposium
Advances in MetaAnalysis for Multivariable Linear Models
Sunday, May 24, 2009,
9:00 AM  10:20 AM
Yerba Buena 3  4
Chair:
Adam R. Hafdahl
Washington University in St. Louis

Most research synthesists metaanalyze one or more separate bivariate associations, each quantified by a correlation or other index. We intend for this symposium to familiarize a broader audience with contemporary metaanalytic techniques better suited for forms of multivariable association that primary researchers often examine: Linear models such as path, factor, and structural equation models, and functions of correlations such as partial and multiple correlations and regression coefficients. Some familiarity with basic metaanalysis and linearmodel concepts and procedures will be assumed, but the presentations are meant to be accessible to the typical psychological scientist.
Demonstration of FixedEffects Model Alternatives for Synthesizing Correlation and Covariance Matrices
S Natasha Beretvas
University of Texas at Austin
Metaanalytic structural equation modeling (MASEM) involves synthesizing elements of correlation or covariance matrices to be used to estimate a structural equation model. Several methods have been suggested for conducting MASEM studies. Using a real dataset, these different methods will be demonstrated. Issues and dilemmas as well as directions for future research will be discussed.
Missing Data in MetaAnalytic Structural Equation Modeling
Carolyn F. Furlow
Georgia State University
In metaanalytic structural equation modeling (MASEM) each primary study contributes a matrix of correlations. Because the research question of interest to the MASEM researcher is frequently different than those of the authors of the primary studies, there are typically variables of interest that are not measured in the primary studies. This results in correlations that are missing in the matrix. This presentation will focus on the various methods for handling those missing correlations as well as some of the issues surrounding the choice of method for handling missing data for MASEM.
MetaAnalysis for Functions of Dependent Correlations
Adam R. Hafdahl
Washington University in St. Louis
Many metaanalysts who work with correlations among several variables are interested in some function of dependent correlations, such as partial or (squared) multiple correlations, regression or path coefficients, or combinations of these quantities. This presentation will cover various metaanalytic techniques for such functions, including illustrations using real data. These procedures comprise both fixed and randomeffects methods, with emphasis on the latter for heterogeneous correlation matrices, as well as inference via largesample and bootstrap strategies.
Modeling Multivariate Effect Sizes With Structural Equation Models
Mike W.L. Cheung
National University of Singapore
Multivariate metaanalysis has become increasingly popular in medical, behavioral and social sciences as the outcome measures in a metaanalysis may involve more than one effect size. Structural equation models (SEMs), on the other hand, provide a flexible statistical framework to model multivariate data. This presentation proposes a new approach to model multivariate effect sizes. Fixed, random and mixedeffects metaanalyses can be analyzed as SEMs. The SEM approach provides an attractive alternative to conducting multivariate metaanalysis.
Go back