Research Methods in Applied Linguistics 1 (2022) 100018
Contents lists available at ScienceDirect
Research Methods in Applied Linguistics
journal homepage: www.elsevier.com/locate/rmal
Applying Meta-Analytic Structural Equation Modeling to Second
Language Research: An Introduction
Amin Raeisi-Vanani a,1 , Luke Plonsky b , Wei Wang c,2,* , Kejin Lee d , Peng Peng e
a Shahid Beheshti University, Tehran, Iran
b Northern Arizona University
c The Graduate Center, City University ofNew York
d The University ofIllinois at Chicago College ofMedicine
e The University ofTexas, Austin
a r t i c l e i n f o
Keywords:
Meta-analysis
Structural equation modeling (SEM)
MASEM
L2 research
Quantitative research methods
a b s t r a c t
Structural equation modeling (SEM) and meta-analysis (MA) are both powerful techniques em-
ployed frequently throughout the social and behavioral sciences, including applied linguistics.
Although meta-analytic data are typically analyzed by calculating weighted means or correlation
coef f i cients, other statistical models such as SEM can also be applied (Schoemann, 2016). SEM
models gauge conceptualized models vis-à-vis empirical data across a given domain. Despite a
considerable expansion of the analytical repertoire in applied linguistics in recent years (Gass,
Loewen, & Plonsky, 2021), this particular technique has yet to be formally introduced or applied.
The present methods tutorial, therefore, aims to introduce MASEM to applied linguistics. In do-
ing so, we provide a conceptual rationale for MASEM, an outline of major stages involved, and
a worked example of how MASEM might be utilized in the f i eld, along with the data and code
necessary for re-running all analyses.
Background
Classical statistical techniques such as multiple regression and analysis of variance (ANOVA) have stood the test of time as
workhorses for many a discipline including applied linguistics (see Plonsky & Oswald, 2017). However, as the f i eld’s research has
matured, leading to more nuanced questions, a move to more cumulatively informed, model-based, and multivariate techniques
has arisen (Hancock & Schoonen, 2015; Larsson, Plonsky & Hancock, in press; Winke, 2014). To be sure, applied linguistics has
expanded its methodological potential in recent years by adopting a more varied, cogent, and sophisticated range of analytical
approaches (Gass et al., 2021). Techniques either introduced or re-considered within this “methodological turn” (Byrnes, 2013, p.
825) include factor analysis (Plonsky & Gonulal, 2015), Bayesian analysis (Norouzian, de Miranda, & Plonsky, 2018), mixed-ef f ects
modeling (Gries, 2021), and cluster analysis (Crowther et al., 2021), amongst others. Two additional techniques now part of the
f i eld’s statistical repertoire include structural equation modeling (SEM) and meta-analysis (e.g., In’nami & Koizumi, 2011; Plonsky &
Oswald, 2015).
*
Corresponding author at: Dr. Wei Wang, CUNY-Graduate Center ZGM: CUNY The Graduate Center, 365 5th Ave, New York, NY 10016, United
States. https://orcid.org/0000-0002-4090-4712.
E-mail address: wwang@gc.cuny.edu (W. Wang).
1
ORCID: http://orcid.org/0000-0002-9575-3165.
2
ORCID: https://orcid.org/0000-0002-4090-4712.
https://doi.org/10.1016/j.rmal.2022.100018
Received 29 December 2021; Received in revised form 1 June 2022; Accepted 2 June 2022
2772-7661/(c) 2022 Elsevier Ltd. All rights reserved.