Search

A growing number of publications focus on estimating Gaussian graphical models (GGM, networks of partial correlation coefficients). At the same time, generalizibility and replicability of these highly parameterized models are debated, and sample sizes typically found in datasets may not be sufficient for estimating the underlying network structure. In addition, while recent work emerged that aims to compare networks based on different samples, these studies do not take potential cross-study heterogeneity into account. To this end, this paper introduces methods for estimating GGMs by aggregating over multiple datasets. We first introduce a general maximum likelihood estimation modeling framework in which all discussed models are embedded. This modeling framework is subsequently used to introduce meta-analytic Gaussian network aggregation (MAGNA). We discuss two variants: fixed-effects MAGNA, in which heterogeneity across studies is not taken into account, and random-effects MAGNA, which models sample correlations and takes heterogeneity into account. We assess the performance of MAGNA in large-scale simulation studies. Finally, we exemplify the method using four datasets of post-traumatic stress disorder (PTSD) symptoms, and summarize findings from a larger meta-analysis of PTSD symptom.

Redundancy analysis (RA) is a multivariate method that maximizes the mean variance of a set of criterion variables explained by a small number of redundancy variates (i.e., linear combinations of a set of predictor variables). However, two challenges exist in RA. First, inferential information for the RA estimates might not be readily available. Second, the existing methods addressing the dimensionality problem in RA are limited for various reasons. To aid the applications of RA, we propose a direct covariance structure modeling approach to RA. The proposed approach (1) provides inferential information for the RA estimates, and (2) allows the researcher to use a simple yet practical criterion to address the dimensionality problem in RA. We illustrate our approach with an artificial example, validate some standard error estimates by simulations, and demonstrate our new criterion in a real example. Finally, we conclude with future research topics.

In this paper, we show how the methods of systematic reviewing and meta-analysis can be used in conjunction with structural equation modeling to summarize the results of studies in a way that will facilitate the theory development and testing needed to advance prevention science. We begin with a high-level overview of the considerations that researchers need to address when using meta-analytic structural equation modeling (MASEM) and then discuss a research project that brings together theoretically important cognitive constructs related to depression to (a) show how these constructs are related, (b) test the direct and indirect effects of dysfunctional attitudes on depression, and (c) test the effects of study-level moderating variables. Our results suggest that the indirect effect of dysfunctional attitudes (via negative automatic thinking) on depression is two and a half times larger than the direct effect of dysfunctional attitudes on depression. Of the three study-level moderators tested, only sample recruitment method (clinical vs general vs mixed) yielded different patterns of results. The primary difference observed was that the dysfunctional attitudes → automatic thoughts path was less strong for clinical samples than it was for general and mixed samples. These results illustrate how MASEM can be used to compare theoretically derived models and predictions resulting in a richer understanding of both the empirical results and the theories underlying them.

Missing data is a common occurrence in confirmatory factor analysis (CFA). Much work had evaluated the performance of different techniques when all observed variables were either continuous or ordinal. However, few have investigated these techniques when observed variables are a mix of continuous and ordinal variables. This study investigated the performance of four approaches to handling missing data in these models: a joint ordinal-continuous full information maximum likelihood (FIML) approach and three multiple imputation approaches (fully conditional specification, fully conditional specification with latent variable formulation, and expectation-maximization with bootstrapping) combined with the weighted least squares with mean and variance adjustment (WLSMV) estimator. In a Monte-Carlo simulation, the FIML approach produced unbiased estimations of factor loadings and standard errors in almost all conditions. Fully conditional specification combined with WLSMV was second best, producing accurate estimates when the sample size was large. However, FIML encountered slight non-convergence issues when certain ordinal categories have extremely low frequencies, which is typical of skewed data. If the sample is large, fully conditional specification combined with weighted least squares is recommended when the FIML approach is not feasible (e.g., non-convergence, impractical computation durations, and variables that predict missingness are not of interest to the analysis).

This meta-analytic study examined average effect sizes across studies for the association between dispositional mindfulness and gratitude, dispositional mindfulness and emotional intelligence, and emotional intelligence and gratitude. The study also tested a model positing emotional intelligence as a path linking mindfulness and gratitude. Two-stage structural equation modeling provided information regarding the association between mindfulness and gratitude across eight samples, the association between mindfulness and emotional intelligence across twenty-six samples, the association between emotional intelligence and gratitude across seven samples, and the mediating role of emotional intelligence in the association between mindfulness and gratitude. Samples were heterogeneous with varying populations. Across samples with a total of 3130 participants, greater mindfulness was significantly associated with more gratitude; r = .22, 95% CI [.16, .28], p

Meta‐analytic structural equation modeling (MASEM) refers to fitting structural equation models (SEMs) (such as path models or factor models) to meta‐analytic data. Currently, fitting MASEMs may be challenging for researchers that are not accustomed to working with R software and packages. Therefore, we developed webMASEM; a web application for MASEM. This app implements the one‐stage MASEM approach, and allows users to apply MASEM in a user‐friendly way. The aim of this article is to provide a tutorial on one‐stage MASEM and a practical guide to webMASEM. We will pay specific attention to how the data should be structured and prepared for webMASEM, because mistakes in this step may lead to faulty results without receiving an error message. The use of webMASEM is illustrated with an analysis of a meta‐analytic path model in which the path coefficients are moderated by a study‐level variable, a meta‐analytic factor model in which the factor loadings are moderated by a study‐level variable, and a meta‐analytic panel model in which the effects are moderated by a study‐level variable. All used datafiles and R scripts are available online.

A recent meta-analytical paper (Bierwiaczonek & Kunst, 2021) sparked a controversy by demonstrating that cross-sectional associations between acculturation and adaptation are weak and heterogenous, and approach zero when assessed over time. Some responses criticized the paper by arguing that small but robust effects can make a real-life difference if applied to millions of migrants. In this paper, we present additional analyses to illustrate how the heterogeneity of correlational acculturation effects complicates and potentially undermines this notion. We estimate that true effects of integration and heritage culture orientation are negative almost 30% of the time, and only true effects of mainstream culture orientation seem to be relatively homogenous. We conclude that the combination of a weak effect with high heterogeneity makes acculturation effects largely unpredictable, meaning that for some migrant groups, adopting the integration strategy will be associated with positive outcomes, and for others with negative outcomes, with no evidence that any of those outcomes are caused by acculturation.

Early childhood is a crucial period for human development that has long-term implications for one’s life trajectories. During the years before formal schooling, brain size and structures, as well as cognitive abilities, undergo rapid development. Children’s cognitive abilities develop by leaps and bounds and show great malleability. Cognitive development during early childhood exerts a long-lasting influence on children’s life chances in adulthood. In China, however, there is no established reliable early childhood achievement test that can be used in a study. This study validates an achievement test for Chinese preschoolers. We analyze data from a nationally representative sample of children aged three to six who participated in the Zhang-Yeung Test of Achievement for Chinese Children (ZY-TACC) in 2012. The instrument consists of a 28-item verbal test and a 24-item numeracy test. Our evaluations indicate a satisfactory level of difficulty, as well as high internal consistency and reliability. This instrument exhibits ample ability to distinguish among children of different ages and varying family backgrounds in ways consistent with previous literature. Children’s test scores are also found to correlate in the expected direction with their behavioral indicators. We demonstrate that the ZY-TACC is a psychometrically robust, culturally and contextually appropriate instrument for assessing Chinese preschool children’s achievement. The instrument can make a significant contribution to research on early childhood development in China.

This quantitative review systematically integrates the antecedents and outcomes of psychological ownership (PO) and examines its incremental validity and explanatory power compared with two other forms of workplace attachment (i.e., organizational commitment and organizational identification). Across 141 studies published over 20 years, our meta-analysis shows that apart from the factors related to the three traditional categories of PO antecedents (i.e., control, knowing, and investment), safety (e.g., organizational justice, trust, perceived organizational support, and relational closeness) is an emerging antecedent leading to PO. In addition, we find that PO is related not only to employees’ attitudinal and performance outcomes but also to some dark-side outcomes (e.g., territorial behaviors). Furthermore, after applying two advanced methods, that is, two-stage meta-analytic structural equation modeling and dominance analysis, to the analysis of 294 studies (including 291 primary studies and three published meta-analyses), the results reveal that PO has an incremental validity above that of organizational commitment and organizational identification in predicting employees’ in-role performance and organizational citizenship behaviors. Theoretical and practical implications are discussed.

Meta-analytic structural equation modeling (MASEM) is a statistical technique to fit hypothesized models on the combined data of multiple independent studies. Lv and Maeda (2019) present a simulation study on the performance of three fixed-effects correlation-based MASEM methods with varying levels of data missing completely at random (MCAR). In this commentary, we discuss several coding errors and other issues that we identified, which demonstrate that Lv and Maeda did not evaluate any of the three intended methods. Furthermore, the authors report very surprising results and offer specific recommendations for the application of the three methods; these actions compel us to express our concerns regarding the validity of the conclusions provided by Lv and Maeda.

Objective: According to the theory of planned behavior, individuals are more likely to act on their behavioral intentions, and report intentions aligned with their attitudes and subjective norm, when their perceived behavioral control (PBC) is high. We tested these predictions meta-analytically by estimating the moderating effect of PBC on the attitude-intention, subjective norm-intention, and the intention-behavior relations in studies applying the theory in the health behavior domain. Method: We conducted a preregistered secondary analysis of studies (k = 39, total N = 13,121) from two programs of research. Each study measured participants’ attitude, subjective norms, PBC, and intentions in relation to health behaviors, and most (k = 36) measured health behavior at follow-up. Data were analyzed using meta-analytic structural equation modeling. Behavior type, scale score coverage, sample age, and publication states were included as moderators of model effects. Results: PBC moderated the intention-behavior relation but not the attitude-intention and subjective norm-intention relations. All moderation effects exhibited significant heterogeneity. Analysis of moderators indicated that the PBC moderation effects on intention varied according to scale score coverage but not by the other moderator variables tested. Conclusions: Results support moderation of the intention-behavior relation by PBC in health behaviors. However, substantial unresolved heterogeneity in the effect across studies remained. Further, these effects may not generalize to other populations and moderator analyses were confined to broad categories. More research that tests these moderation effects in health behavior contexts and reports sufficient data necessary for conducting a meta-analysis is needed to enable moderator analyses with greater fidelity. peerReviewed

A growing number of publications focus on estimating Gaussian graphical models (GGM, networks of partial correlation coefficients). At the same time, generalizibility and replicability of these highly parameterized models are debated, and sample sizes typically found in datasets may not be sufficient for estimating the underlying network structure. In addition, while recent work emerged that aims to compare networks based on different samples, these studies do not take potential cross-study heterogeneity into account. To this end, this paper introduces methods for estimating GGMs by aggregating over multiple datasets. We first introduce a general maximum likelihood estimation modeling framework in which all discussed models are embedded. This modeling framework is subsequently used to introduce meta-analytic Gaussian network aggregation (MAGNA). We discuss two variants: fixed-effects MAGNA, in which heterogeneity across studies is not taken into account, and random-effects MAGNA, which models sample correlations and takes heterogeneity into account. We assess the performance of MAGNA in large-scale simulation studies. Finally, we exemplify the method using four datasets of post-traumatic stress disorder (PTSD) symptoms, and summarize findings from a larger meta-analysis of PTSD symptom.

Machine learning methods have become very popular in diverse fields due to their focus on predictive accuracy, but little work has been conducted on how to assess the replicability of their findings. We introduce and adapt replication methods advocated in psychology to the aims and procedural needs of machine learning research. In Study 1, we illustrate these methods with the use of an empirical data set, assessing the replication success of a predictive accuracy measure, namely, R2 on the cross-validated and test sets of the samples. We introduce three replication aims. First, tests of inconsistency examine whether single replications have successfully rejected the original study. Rejection will be supported if the 95% confidence interval (CI) of R2 difference estimates between replication and original does not contain zero. Second, tests of consistency help support claims of successful replication. We can decide apriori on a region of equivalence, where population values of the difference estimates are considered equivalent for substantive reasons. The 90% CI of a different estimate lying fully within this region supports replication. Third, we show how to combine replications to construct meta-analytic intervals for better precision of predictive accuracy measures. In Study 2, R2 is reduced from the original in a subset of replication studies to examine the ability of the replication procedures to distinguish true replications from nonreplications. We find that when combining studies sampled from same population to form meta-analytic intervals, random-effects methods perform best for cross-validated measures while fixed-effects methods work best for test measures. Among machine learning methods, regression was comparable to many complex methods, while support vector machine performed most reliably across a variety of scenarios. Social scientists who use machine learning to model empirical data can use these methods to enhance the reliability of their findings.

Automated magnetic resonance imaging (MRI) volumetry is a promising tool to evaluate regional brain volumes in dementia and especially Alzheimer's disease (AD).To compare automated methods and the gold standard manual segmentation in measuring regional brain volumes on MRI across healthy controls, patients with mild cognitive impairment, and patients with dementia due to AD.Systematic review and meta-analysis.MEDLINE, Embase, and PsycINFO were searched through October 2021.1.0 T, 1.5 T, or 3.0 T.Two review authors independently identified studies for inclusion and extracted data. Methodological quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2).Standardized mean differences (SMD; Hedges' g) were pooled using random-effects meta-analysis with robust variance estimation. Subgroup analyses were undertaken to explore potential sources of heterogeneity. Sensitivity analyses were conducted to examine the impact of the within-study correlation between effect estimates on the meta-analysis results.Seventeen studies provided sufficient data to evaluate the hippocampus, lateral ventricles, and parahippocampal gyrus. The pooled SMD for the hippocampus, lateral ventricles, and parahippocampal gyrus were 0.22 (95% CI -0.50 to 0.93), 0.12 (95% CI -0.13 to 0.37), and -0.48 (95% CI -1.37 to 0.41), respectively. For the hippocampal data, subgroup analyses suggested that the pooled SMD was invariant across clinical diagnosis and field strength. Subgroup analyses could not be conducted on the lateral ventricles data and the parahippocampal gyrus data due to insufficient data. The results were robust to the selected within-study correlation value.While automated methods are generally comparable to manual segmentation for measuring hippocampal, lateral ventricle, and parahippocampal gyrus volumes, wide 95% CIs and large heterogeneity suggest that there is substantial uncontrolled variance. Thus, automated methods may be used to measure these regions in patients with AD but should be used with caution.3 TECHNICAL EFFICACY: Stage 3.

In this paper, we show how the methods of systematic reviewing and meta-analysis can be used in conjunction with structural equation modeling to summarize the results of studies in a way that will facilitate the theory development and testing needed to advance prevention science. We begin with a high-level overview of the considerations that researchers need to address when using meta-analytic structural equation modeling (MASEM) and then discuss a research project that brings together theoretically important cognitive constructs related to depression to (a) show how these constructs are related, (b) test the direct and indirect effects of dysfunctional attitudes on depression, and (c) test the effects of study-level moderating variables. Our results suggest that the indirect effect of dysfunctional attitudes (via negative automatic thinking) on depression is two and a half times larger than the direct effect of dysfunctional attitudes on depression. Of the three study-level moderators tested, only sample recruitment method (clinical vs general vs mixed) yielded different patterns of results. The primary difference observed was that the dysfunctional attitudes → automatic thoughts path was less strong for clinical samples than it was for general and mixed samples. These results illustrate how MASEM can be used to compare theoretically derived models and predictions resulting in a richer understanding of both the empirical results and the theories underlying them.

For idiopathic generalized epilepsies (IGE), brain network analysis is emerging as a biomarker for potential use in clinical care. To determine whether people with IGE show alterations in resting-state brain connectivity compared to healthy controls, and to quantify these differences, we conducted a systematic review and meta-analysis of EEG and magnetoencephalography (MEG) functional connectivity and network studies. The review was conducted according to PRISMA guidelines. Twenty-two studies were eligible for inclusion. Outcomes from individual studies supported hypotheses for interictal, resting-state brain connectivity alterations in IGE patients compared to healthy controls. In contrast, meta-analysis from six studies of common network metrics clustering coefficient, path length, mean degree and nodal strength showed no significant differences between IGE and control groups (effect sizes ranged from -0.151 -1.78). The null findings of the meta-analysis and the heterogeneity of the included studies highlights the importance of developing standardized, validated methodologies for future research. Network neuroscience has significant potential as both a diagnostic and prognostic biomarker in epilepsy, though individual variability in network dynamics needs to be better understood and accounted for.

The mathematical connection between canonical correlation analysis (CCA) and covariance structure analysis was first discussed through the Multiple Indicators and Multiple Causes (MIMIC) approach. However, the MIMIC approach has several technical and practical challenges. To address these challenges, a comprehensive COSAN modeling approach is proposed. Specifically, we define four COSAN-CCA models to correspond with four possible combinations of the data to be analyzed and the unique parameters to be estimated. In terms of the data, one can analyze either the unstandardized or standardized variables. In terms of the unique parameters, one can estimate either the weights or loadings. Besides the unique parameters of each COSAN-CCA model, all four COSAN-CCA models also estimate the canonical correlations as their common parameters. Taken together, the four COSAN-CCA models provide the correct point estimates and standard error estimates for all commonly used CCA parameters. Two numeric examples are used to compare the standard error estimates obtained from the MIMIC approach and the COSAN modeling approach. Moreover, the standard error estimates from the COSAN modeling approach are validated by a simulation study and the asymptotic theory. Finally, software implementation and future extensions are discussed.

This study investigates the mechanism between ethical leadership and employees’ organizational citizenship behavior (OCB) through a justice perspective. We propose that interactional justice serves as a conduit that induces employees’ OCB in response to leaders’ ethical behaviors. The explanatory powers of interactional justice on two forms of OCB – organizational OCB (OCBO) and interpersonal OCB (OCBI) – are compared. Based on data from 34,028 participants in 100 empirical studies, we applied a two-stage meta-analytic structural equations modeling (TSSEM) and found that interactional justice only explains the influence of ethical leadership on employees’ OCBO but not OCBI. Theoretical and managerial implications of our findings are discussed.

Structural equation modeling (SEM) and meta-analysis are two popular techniques in the behavioral, medical, and social sciences. They have their own research communities, terminologies, models, software packages, and even journals. This chapter introduces SEM-based meta-analysis, an approach to conduct meta-analyses using the SEM framework. By conceptualizing studies in a meta-analysis as subjects in a structural equation model, univariate, multivariate, and three-level meta-analyses can be fitted as structural equation models using definition variables. We will review fixed-, random-, and mixed-effects models using the SEM framework. Examples will be used to illustrate the procedures using the metaSEM and OpenMx packages in R. This chapter closes with a discussion of some future directions for research.

Missing data is a common occurrence in confirmatory factor analysis (CFA). Much work had evaluated the performance of different techniques when all observed variables were either continuous or ordinal. However, few have investigated these techniques when observed variables are a mix of continuous and ordinal variables. This study investigated the performance of four approaches to handling missing data in these models: a joint ordinal-continuous full information maximum likelihood (FIML) approach and three multiple imputation approaches (fully conditional specification, fully conditional specification with latent variable formulation, and expectation-maximization with bootstrapping) combined with the weighted least squares with mean and variance adjustment (WLSMV) estimator. In a Monte-Carlo simulation, the FIML approach produced unbiased estimations of factor loadings and standard errors in almost all conditions. Fully conditional specification combined with WLSMV was second best, producing accurate estimates when the sample size was large. However, FIML encountered slight non-convergence issues when certain ordinal categories have extremely low frequencies, which is typical of skewed data. If the sample is large, fully conditional specification combined with weighted least squares is recommended when the FIML approach is not feasible (e.g., non-convergence, impractical computation durations, and variables that predict missingness are not of interest to the analysis).

Posttraumatic stress disorder (PTSD) researchers have increasingly used psychological network models to investigate PTSD symptom interactions, as well as to identify central driver symptoms. It is unclear, however, how generalizable such results are. We have developed a meta-analytic framework for aggregating network studies while taking between-study heterogeneity into account and applied this framework in the first-ever meta-analytic study of PTSD symptom networks. We analyzed the correlational structures of 52 different samples with a total sample size of n = 29,561 and estimated a single pooled network model underlying the data sets, investigated the scope of between-study heterogeneity, and assessed the performance of network models estimated from single studies. Our main findings are that: (a) We identified large between-study heterogeneity, indicating that it should be expected for networks of single studies to not perfectly align with one-another, and meta-analytic approaches are vital for the study of PTSD networks. (b) While several clear symptom-links, interpretable clusters, and significant differences between strength of edges and centrality of nodes can be identified in the network, no single or small set of nodes that clearly played a more central role than other nodes could be pinpointed, except for the symptom "amnesia" that was clearly the least central symptom. (c) Despite large between-study heterogeneity, we found that network models estimated from single samples can lead to similar network structures as the pooled network model. We discuss the implications of these findings for both the PTSD literature as well as methodological literature on network psychometrics. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Meta-analytic structural equation modeling (MASEM) is an increasingly popular meta-analytic technique that combines the strengths of meta-analysis and structural equation modeling. MASEM facilitates the evaluation of complete theoretical models (e.g., path models or factor analytic models), accounts for sampling covariance between effect sizes, and provides measures of overall fit of the hypothesized model on meta-analytic data. We propose a novel MASEM method, one-stage MASEM, which is better suitable to explain study-level heterogeneity than existing methods. One-stage MASEM allows researchers to incorporate continuous or categorical moderators into the MASEM, in which any parameter in the structural equation model (e.g., path coefficients and factor loadings) can be modeled by the moderator variable, while the method does not require complete data for the primary studies included in the meta-analysis. We illustrate the new method on two real data sets, evaluate its empirical performance via a computer simulation study, and provide user-friendly R-functions and annotated syntax to assist researchers in applying one-stage MASEM. We close the article by presenting several future research directions. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

AimsA mediator is a variable that explains the underlying mechanism between an independent variable and a dependent variable. The indirect effect indicates the effect from the predictor to the outcome variable via the mediator. In contrast, the direct effect represents the predictor's effort on the outcome variable after controlling for the mediator.MethodsA single study rarely provides enough evidence to answer research questions in a particular domain. Replications are generally recommended as the gold standard to conduct scientific research. When a sufficient number of studies have been conducted addressing similar research questions, a meta-analysis can be used to synthesize those studies' findings.ResultsThe main objective of this paper is to introduce two frameworks to integrating studies using mediation analysis. The first framework involves calculating standardized indirect effects and direct effects and conducting a multivariate meta-analysis on those effect sizes. The second one uses meta-analytic structural equation modeling to synthesize correlation matrices and fit mediation models on the average correlation matrix. We illustrate these procedures on a real dataset using the R statistical platform.ConclusionThis paper closes with some further directions for future studies.

Meta-analytic Structural Equation Modeling (MASEM) refers to fitting structural equation models (such as path models or factor models) to meta-analytic data. Currently, fitting MASEMs may be challenging for researchers that are not accustomed to working with R software and packages. Therefore, we developed webMASEM; a web application for MASEM. This app implements the one-stage MASEM approach, and allows users to apply MASEM in a user-friendly way. The aim of this article is to provide a tutorial on one-stage MASEM and a practical guide to webMASEM. We will pay specific attention to how the data should be structured and prepared for webMASEM, because mistakes in this step may lead to faulty results without receiving an error message. The use of webMASEM is illustrated with an analysis of a meta-analytic path model in which the path coefficients are moderated by a study-level variable, and a meta-analytic factor model in which the factor loadings are moderated by a study-level variable. All used datafiles and R-codes are available online.

Internet gaming disorder (IGD) has been viewed by scholars as (a) a pathology that co-occurs with psychological problems (comorbidity hypothesis), (b) maladaptive coping with abundant interpersonal problems (interpersonal impairment hypothesis), and (c) deficient self-regulation with the underlying motive to restore psychosocial well-being (dilution effect hypothesis). We examined the associations between IGD symptoms and four major criteria (psychological problems, interpersonal problems, psychological well-being, and interpersonal well-being), and compared the magnitude of these associations across countries. To test these hypotheses, we performed mixed-effects meta-analysis on 84 independent samples comprising 58,834 participants from 20 countries. The findings showed moderately strong positive associations between IGD symptoms and psychological problems across the countries, providing some support for the universality of the comorbidity hypothesis. The interpersonal impairment hypothesis was more tenable to countries lower (vs. higher) in power distance, which exhibited a stronger (vs. weaker) positive correlation between IGD symptoms and interpersonal problems. The dilution effect hypothesis was more tenable to countries either higher (vs. lower) in national life satisfaction or lower (vs. higher) in cultural masculinity, each of which displayed a weaker (vs. stronger) inverse correlation between IGD symptoms and interpersonal well-being.

Abstract. Machine learning tools are increasingly used in social sciences and policy fields due to their increase in predictive accuracy. However, little research has been done on how well the models of machine learning methods replicate across samples. We compare machine learning methods with regression on the replicability of variable selection, along with predictive accuracy, using an empirical dataset as well as simulated data with additive, interaction, and non-linear squared terms added as predictors. Methods analyzed include support vector machines (SVM), random forests (RF), multivariate adaptive regression splines (MARS), and the regularized regression variants, least absolute shrinkage and selection operator (LASSO), and elastic net. In simulations with additive and linear interactions, machine learning methods performed similarly to regression in replicating predictors; they also performed mostly equal or below regression on measures of predictive accuracy. In simulations with square terms, machine learning methods SVM, RF, and MARS improved predictive accuracy and replicated predictors better than regression. Thus, in simulated datasets, the gap between machine learning methods and regression on predictive measures foreshadowed the gap in variable selection. In replications on the empirical dataset, however, improved prediction by machine learning methods was not accompanied by a visible improvement in replicability in variable selection. This disparity is explained by the overall explanatory power of the models. When predictors have small effects and noise predominates, improved global measures of prediction in a sample by machine learning methods may not lead to the robust selection of predictors; thus, in the presence of weak predictors and noise, regression remains a useful tool for model building and replication.

Meta-analysis and structural equation modeling (SEM) are 2 of the most prominent statistical techniques employed in the behavioral, medical, and social sciences. They each have their own well-established research communities, terminologies, statistical models, software packages, and journals (Research Synthesis Methods and Structural Equation Modeling: A Multidisciplinary Journal). In this paper, I will provide some personal reflections on combining meta-analysis and SEM in the forms of meta-analytic SEM and SEM-based meta-analysis. The critical contributions of Becker (1992), Shadish (1992), and Viswesvaran and Ones (1995) in the early development of meta-analytic SEM are highlighted. Another goal of the paper is to illustrate how meta-analysis can be extended and integrated with other techniques to address new research questions such as the analysis of Big Data. I hope that this paper may stimulate more research development in the area of combining meta-analysis and SEM.

Meta-analytic structural equation modeling (MASEM) is becoming increasingly popular for testing theoretical models from a pool of correlation matrices in management and organizational studies. One limitation of the conventional MASEM approaches is that the proposed structural equation models are only tested on the average correlation matrix. It remains unclear how far the proposed models can be generalized to other populations when the correlation matrices are heterogeneous. Recently, Yu, Downes, Carter, and O'Boyle (2016) proposed a full-information MASEM approach to address this limitation by fitting structural equation models from the correlation matrices generated from a parametric bootstrap. However, their approach suffers from several conceptual issues and technical errors. In this study, we reran some of the simulations in Yu et al. by correcting all of the errors in their original studies. The findings showed that bootstrap credible intervals (CVs) work reasonably well, whereas test statistics and goodness-of-fit indices do not. We advise researchers on what they can and cannot achieve by applying the full-information MASEM approach. We recommend fitting MASEM with the two-stage structural equation modeling approach, which works well for the simulation studies. If researchers want to inspect the heterogeneity of the parameters, they may use the bootstrap CVs from the full-information MASEM approach. All of these analyses were implemented in the open-source R statistical platform; researchers can easily apply and verify the findings. This article concludes with several future directions to address the issue of heterogeneity in MASEM. (PsycINFO Database Record

Meta-analytic structural equation modeling (MASEM) is a statistical technique to pool correlation matrices and test structural equation models on the pooled correlation matrix. In Stage 1 of MASEM, correlation matrices from independent studies are combined to obtain a pooled correlation matrix, using fixed- or random-effects analysis. In Stage 2, a structural model is fitted to the pooled correlation matrix. Researchers applying MASEM may have hypotheses about how certain model parameters will differ across subgroups of studies. These moderator hypotheses are often addressed using suboptimal methods. The aim of the current article is to provide guidance and examples on how to test hypotheses about group differences in specific model parameters in MASEM. We illustrate the procedure using both fixed- and random-effects subgroup analysis with two real datasets. In addition, we present a small simulation study to evaluate the effect of the number of studies per subgroup on convergence problems. All data and the R-scripts for the examples are provided online.

The current study examines the relationship between bilingual children's dual language experience (i.e. language input, language output and vocabulary proficiency), and their social-emotional and b...

A number of novel problematic behaviors have emerged in the information technology era, and corresponding addictions have been proposed for some of these behaviors. Scholars have speculated that a common factor may underlie these information technology addictions, but this theoretical notion has yet to be tested empirically. The present study tested this notion and also investigated the relationships of information technology addictions with other behavioral addictions as well as substance addictions. We conducted an online survey in 1001 US adults (56% female; mean age = 35.0 years, range = 18–83). Two conceptual models were formulated and tested. Moreover, correlations of the information technology addictions with both problematic gambling and alcohol use disorder were examined. The confirmatory factor analysis showed that there was a common factor underlying various types of information technology addiction. In addition, problematic gambling was more strongly correlated with information technology addiction than alcohol use disorder was. Our findings are interpreted in light of a spectrum approach, which conceptualizes information technology addiction as a cluster of disorders comprising not only shared risk factors and symptoms but also distinct characteristics. The findings further reveal that information technology addiction is more similar to other behavioral addictions than substance-related addictions. Implications for researchers and practitioners are discussed.

Statistical methods play an important role in behavioural, medical, and social sciences. Two recent statistical advances are structural equation modelling (SEM) and meta-analysis. SEM is used to test hypothesised models based on substantive theories, which can be path, confirmatory factor analytic, or full structural equation models. Meta-analysis is used to synthesise research findings in a particular topic. This article demonstrates another recent statistical advance - meta-analytic structural equation modelling (MASEM) - that combines meta-analysis and SEM to synthesise research findings for the purpose of testing hypothesised models. Using the theory of planned behaviour as an example, we show how MASEM can be used to address important research questions that cannot be answered by univariate meta-analyses on Pearson correlations. Specifically, MASEM allows researchers to: (1) test whether the proposed models are consistent with the data; (2) estimate partial effects after controlling for other variables; (3) estimate functions of parameter estimates such as indirect effects; and (4) include latent variables in the models. We illustrate the procedures with an example on the theory of planned behaviour. Practical issues in MASEM and suggested solutions are discussed.

ObjectiveThe association between cigarette smoking and the risk of systemic lupus erythematosus (SLE) remains a matter for debate. Additionally, the effect of the change of smokers’ demographics on the risk of development of SLE over time has not been formally addressed. We aimed to examine the association between cigarette smoking and the risk of SLE by performing an updated metaanalysis.MethodsA literature search using keywords including “lupus,” “smoking,” “cigarette,” “environmental,” “autoimmune,” and “connective tissue disease” was performed in computerized databases to identify studies addressing the relationship between cigarette smoking and SLE occurrence. A Bayesian metaanalysis was conducted by computing the log-OR between current and never smokers, and between former and never smokers. The average log-OR (subsequently converted to OR) and their corresponding 95% credible intervals (CrI) were calculated. The effect of publication time, sex, and age of patients with SLE on the effect sizes was examined by multivariate metaregression.ResultsData aggregation of 12 eligible studies comprising 3234 individuals who developed SLE and 288,336 control subjects revealed a significant association between SLE occurrence and current smoking status (OR 1.54, 95% CrI 1.06–2.25), while only a non-significant trend was demonstrated between SLE occurrence and former smoking status (OR 1.39, 95% CrI 0.95–2.08). Publication time, sex, and the mean age of patients with SLE did not explain the heterogeneity of the effect sizes.ConclusionCurrent smoking status is associated with risk of SLE. Sex and age of patients with SLE had no significant effect on the risk of SLE over time.

Purpose The purpose of this paper is to illustrate how international business (IB) researchers can benefit from meta-analytic structural equation modeling (MASEM) by introducing a statistically rigorous approach (i.e. two-stage meta-analytic structural equation modeling or TSSEM) and comparing it with a conventional approach (i.e. the univariate-r approach). The illustration and comparison present a methodological overview of MASEM that will assist IB researchers in selecting an optimal method. Design/methodology/approach In this paper, the MASEM method is elaborated upon, and methodological issues are addressed, by comparing the TSSEM and the univariate-r approaches using an empirical illustration. In this illustrative example, which is based on transaction cost economics, the effects of a firm’s internal factors on its levels of commitment in an international entry strategy are examined. Findings The MASEM method can help IB researchers to test and build on IB theories by synthesizing findings in the extant literature because this method reflects the theoretical complexity of IB (e.g. intercorrelationships among factors). Comparing the two approaches of MASEM, it is found in this study that due to its statistical rigorousness TSSEM has methodological advantages in helping IB researchers test theoretical models. Originality/value This is the first study to introduce MASEM into the discipline of IB strategies. In this paper, the authors introduce an advanced research method and illustrate two ways of using it.

Health locus of control (HLOC) refers to beliefs regarding how one’s health is influenced by oneself, others, or fate. This meta-analysis investigated whether three HLOC dimensions (internality/I-HLOC, powerful others/P-HLOC, chance/C-HLOC) were related to both specific health behaviours and global health appraisal, and whether these relationships were moderated by gender and age compositions, individualism, and power distance. Three-level mixed-effects meta-analysis was performed on studies examining the associations of HLOC with specific health behaviour (k = 76, N = 76,580, 57% women, M age = 43.75) and global health appraisal (k = 95, N = 12,068, 57% women, M age = 45.44), respectively. For specific health behaviour, the averaged correlations with the HLOC dimensions were generally weak (r’s = −.07 to .10). However, the links between P-HLOC and exercise were moderated by all four demographic moderators, and gender composition and individualism moderated the association between the HLOC dimensions and diet. For global health appraisal, all of the averaged correlations were statistically significant (r’s = −.16 to .21), except that between P-HLOC and mental quality of life. The results further showed individualism and power distance to moderate the links between the HLOC dimensions and both mental and physical quality of life, and gender composition to moderate those between these dimensions and two indicators of emotional problems (depression and anxiety).

Meta-analytic structural equation modeling (MASEM) combines the techniques of meta-analysis and structural equation modeling for the purpose of synthesizing correlation or covariance matrices and fitting structural equation models on the pooled correlation or covariance matrix. Both fixed-effects and random-effects models can be defined in MASEM. Random-effects models are well known in conventional meta-analysis but are less studied in MASEM. The primary objective of this paper was to address issues related to random-effects models in MASEM. Specifically, we compared two different random-effects models in MASEM-correlation-based MASEM and parameter-based MASEM-and explored their strengths and limitations. Two examples were used to illustrate the similarities and differences between these models. We offered some practical guidelines for choosing between these two models. Future directions for research on random-effects models in MASEM were also discussed. Copyright © 2016 John Wiley & Sons, Ltd.

This study used the latent profile transition analysis (LPTA) to analyze whether emotional labor profiles change across time and how these profiles relate to occupational well-being (i.e., job satisfaction, quality of work life, psychological distress, and work-family conflict). A total of 155 full-time Chinese employees completed the questionnaire survey at two time points. Three latent profiles were identified at Time 1 and the same profiles were replicated at Time 2. We determined that the majority of the participants retained the original profiles. Lastly, occupational well-being differed significantly across the identified profiles. The limitations and implications of this study were also provided.

Meta-analytic structural equation modeling (MASEM) is increasingly applied to advance theories by synthesizing existing findings. MASEM essentially consists of two stages. In Stage 1, a pooled correlation matrix is estimated based on the reported correlation coefficients in the individual studies. In Stage 2, a structural model (such as a path model) is fitted to explain the pooled correlations. Frequently, the individual studies do not provide all the correlation coefficients between the research variables. In this study, we modify the currently optimal MASEM-method to deal with missing correlation coefficients, and compare its performance with existing methods. This study is the first to evaluate the performance of fixed-effects MASEM methods under different levels of missing correlation coefficients. We found that the often used univariate methods performed very poorly, while the multivariate methods performed well overall.