Meta-Analytic Methods of Pooling Correlation Matrices for Structural Equation Modeling Under Different Patterns of Missing Data.

Abstract

Three methods of synthesizing correlations for meta-analytic structural equation modeling (SEM) under different degrees and mechanisms of missingness were compared for the estimation of correlation and SEM parameters and goodness-of-fit indices by using Monte Carlo simulation techniques. A revised generalized least squares (GLS) method for synthesizing correlations, weighted-covariance GLS (W-COV GLS), was compared with univariate weighting with untransformed correlations (univariate r) and univariate weighting with Fisher's z-transformed correlations (univariate z). These 3 methods were crossed with listwise and pairwise deletion. Univariate z and W-COV GLS performed similarly, with W-COV GLS providing slightly better estimation of parameters and more correct model rejection rates. Missing not at random data produced high levels of relative bias in correlation and model parameter estimates and higher incorrect SEM model rejection rates. Pairwise deletion resulted in inflated standard errors for all synthesis methods and higher incorrect rejection rates for the SEM model with univariate weighting procedures.

Safety at work: A meta-analytic investigation of the link between job demands, job resources, burnout, engagement, and safety outcomes.

“…For example, in the current study, the meta-analytic correlation for the relationship between complexity and burnout is obtained from studies that are different from the studies that are utilized to obtain the metaanalytic correlation for the relationship between physical demands and burnout. Research has found parameter estimates and chisquare tests to be biased when data in a meta-analytic matrix are missing not at random (Furlow & Beretvas, 2005). Thus, if effect sizes are missing not at random, the results may be subject to bias (Naragon-Gainey, 2010).…”

“…Third, several limitations in the use of meta-analytic correlation matrices with regard to path analyses have been noted and investigated by researchers (Field, 2005;Furlow & Beretvas, 2005;Hafdahl, 2007;Hafdahl & Williams, 2009). These limitations primarily focus on the accuracy of parameter estimates and goodness-of fit-indices.…”

“…Another concern in using a meta-analytic correlation matrix is that different studies contribute to different cells in the matrix (i.e., sample size varies across cells). Research has found parameter estimates and chi-square tests to be biased when data in a meta-analytic matrix are missing not at random (Furlow & Beretvas, 2005;NaragonGainey, 2010). Potential limitations of meta-analytic correlation matrices are noted in the current study.…”

“…We also utilized the harmonic means of the cell sample sizes according to procedures outlined by Viswesvaran and Ones (1995). Research investigating the use of meta-analytic correlation matrices in several simulation studies has found that parameter estimates remain accurate and that goodness-of-fit indices such as the comparative fit index (CFI) and standardized root mean square residual (SRMR) function well (Beretvas & Furlow, 2006;Cheung & Chan, 2005;Furlow & Beretvas, 2005). Another concern in using a meta-analytic correlation matrix is that different studies contribute to different cells in the matrix (i.e., sample size varies across cells).…”

A model for integrating fixed-, random-, and mixed-effects meta-analyses into structural equation modeling.

“…Besides fitting regression models on the pooled correlation matrix, MASEM has been extended to fit path models, exploratory factor analysis, confirmatory factor analysis, and structural equation models (e.g., Becker, 2000;Becker & Schram, 1994;S. F. Cheung, 2000;Furlow & Beretvas, 2005;Hafdahl, 2001;Shadish, 1996;Viswesvaran & Ones, 1995).…”