AN ANALYSIS OF A NOVEL MASEM APPROACH: DOES OSMASEM MEASURE MISSPECIFICATION CORRECTLY?

Abstract

Meta-Analytic Structural Equation Modelling (MASEM) is a novel approach that combines meta-analysis and SEM together. It enables researchers to take a whole model approach in testing meta-analysed effects while also allowing the hypothesized model to be tested for its fit to the observed data. Current MASEM methods are disadvantaged as they are not able to evaluate both categorical and continuous moderators. In contrast, the one-Step MASEM (OSMASEM) proposed by Jak and Cheung (2019) is a promising novel approach to MASEM that accounts for these flaws. Thus far, initial simulations are promising (Jak & Cheung, 2019). However, it is not yet clear how SEM fit indices perform under OSMASEM or other MASEMs. Thus, to supplement initial findings, the present thesis conducted a series of simulations to test the performance of fit indices from Univariate-R, TSSEM, and OSMASEM and compared them to normal SEM under differing conditions of heterogeneity, model misspecification, model size, and study size. Results indicate that under fixed effects conditions, OSMASEM and TSSEM, behave relatively similarly to normal SEM, while under random effects conditions, OSMASEM and TSSEM performs better than normal SEM. Recommendations and guidelines based on the findings are proposed.