A Primer on Bayesian Model-Averaged Meta-Analysis

Gronau, Quentin Frederik; Heck, Daniel W.; Berkhout, Sophie W.; Haaf, Julia M.; Wagenmakers, Eric–Jan

doi:10.1177/25152459211031256

Cited by 49 publications

(88 citation statements)

References 72 publications

(106 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thereby, multiple mixedeffects models can be used to represent the null and the alternative hypotheses which are of substantive interest. In a simple repeated measures design, we can test for the presence (models M 4 and M 6 ) or absence (models M 3 and M 5 ) of an effect of condition while considering the remaining uncertainty about individual differences (Gronau et al, 2021;Hinne et al, 2020). The inclusion of multiple candidate models in the analysis increases transparency regarding the auxiliary assumptions about the fixed-and random-effects structure (Rouder, Morey, & Wagenmakers, 2016).…”

Section: Discussionmentioning

confidence: 99%

“…Most importantly, the paper shows that researchers should be aware that auxiliary assumptions are required for translating substantive hypotheses to specific statistical models (Kellen, 2019;Suppes, 1966). While Bayesian model selection is ideally suited for testing substantive theories in psychology (Heck et al, 2021), Bayes factors for mixed models necessarily depend on details of the model specification such as the inclusion of fixed and random effects and the prior distributions. Bayesian model averaging allows researchers to make such researchers degrees of freedom transparent by considering multiple model versions at once, thereby accounting for the inherent uncertainty about auxiliary assumptions in mixed-effects modeling.…”

Section: Discussionmentioning

confidence: 99%

“…First, it is necessary to sum the prior and posterior model probabilities for the two model sets as discussed above. Second, the inclusion Bayes factor is obtained by dividing the posterior inclusion odds by the prior inclusion odds (Gronau et al, 2021):…”

Section: Bayesian Model Averagingmentioning

confidence: 99%

“…Instead of using the Bayes factor as a tool for comparing only two specific, nested models, Bayesian model averaging allows researchers to compare a larger set of theoretically interesting models at once. Moreover, the inclusion Bayes factor provides a means for quantifying the amount of evidence for or against the average effect of condition in a repeated measures design while accounting for the uncertainty in model specification (Gronau et al, 2021;Hinne et al, 2020).…”

mentioning

confidence: 99%

See 3 more Smart Citations

Benefits of Bayesian Model Averaging for Mixed-Effects Modeling

Heck

Bockting

2021

Comput Brain Behav

Self Cite

View full text Add to dashboard Cite

Bayes factors allow researchers to test the effects of experimental manipulations in within-subjects designs using mixed-effects models. van Doorn et al. (2021) showed that such hypothesis tests can be performed by comparing different pairs of models which vary in the specification of the fixed- and random-effect structure for the within-subjects factor. To discuss the question of which model comparison is most appropriate, van Doorn et al. compared three corresponding Bayes factors using a case study. We argue that researchers should not only focus on pairwise comparisons of two nested models but rather use Bayesian model selection for the direct comparison of a larger set of mixed models reflecting different auxiliary assumptions regarding the heterogeneity of effect sizes across individuals. In a standard one-factorial, repeated measures design, the comparison should include four mixed-effects models: fixed-effects H0, fixed-effects H1, random-effects H0, and random-effects H1. Thereby, one can test both the average effect of condition and the heterogeneity of effect sizes across individuals. Bayesian model averaging provides an inclusion Bayes factor which quantifies the evidence for or against the presence of an average effect of condition while taking model selection uncertainty about the heterogeneity of individual effects into account. We present a simulation study showing that model averaging among a larger set of mixed models performs well in recovering the true, data-generating model.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Bayesian Model Averagingmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Benefits of Bayesian Model Averaging for Mixed-Effects Modeling

Heck

Bockting

2021

Comput Brain Behav

Self Cite

View full text Add to dashboard Cite

show abstract

“…In Bayesian meta-analysis, model averaging has recently been applied successfully in several applications (Gronau, Van Erp, et al, 2017;Haaf et al, 2020). For statistical details we refer the reader to Gronau et al (2021).…”

Section: Fixed-effects Hmentioning

confidence: 99%

A Tutorial on Bayesian Model-Averaged Meta-Analysis in JASP

Berkhout¹,

Haaf²,

Gronau³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Researchers conduct a meta-analysis in order to synthesize information across different studies. Compared to standard meta-analytic methods, Bayesian model-averaged meta-analysis offers several practical advantages including the ability to quantify evidence in favor of the absence of an effect, the ability to monitor evidence as individual studies accumulate indefinitely, and the ability to draw inferences based on multiple models simultaneously. This tutorial introduces the concepts and logic underlying Bayesian model-averaged meta-analysis and illustrates its application using the open-source software JASP. As a running example, we perform a Bayesian meta-analysis on language development in children. We show how to conduct a Bayesian model-averaged meta-analysis and how to interpret the results.

show abstract

Patent Ductus Arteriosus and Bronchopulmonary Dysplasia–Associated Pulmonary Hypertension

Villamor,

van Westering-Kroon,

Gonzalez-Luis

et al. 2023

JAMA Netw Open

View full text Add to dashboard Cite

ImportanceBronchopulmonary dysplasia (BPD) is often associated with pulmonary vascular disease and secondary pulmonary hypertension (PH). The pathogenesis of BPD-associated PH (BPD-PH) is complex and involves prenatal and postnatal factors that disrupt pulmonary vascular development, and patent ductus arteriosus (PDA) is a factor potentially associated with risk of BPD-PH that has been identified in very recent studies.ObjectiveTo explore the association of PDA with BPD-PH using a bayesian model-averaged (BMA) meta-analysis of studies.Data SourcesPubMed and Embase were searched up to April 2023. Key search terms included BPD and PH.Study SelectionStudies examining infants with gestational age 32 weeks or less and reporting data on PDA and risk of BPD-PH.Data Extraction and SynthesisThis study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses and the Meta-Analysis of Observational Studies in Epidemiology reporting guidelines. Two independent reviewers extracted data, with a third reviewer checking for accuracy and completeness. Data pooling and effect size calculations were performed by BMA.Main Outcomes and MeasuresThe primary outcome was BPD-PH. BMA was used to calculate Bayes factors (BFs). The BF10 is the ratio of the probability of the data under the alternative hypothesis (H1, association of PDA with BPD-HP) over the probability of the data under the null hypothesis (H0).ResultsA total of 32 studies (8513 infants) were included. BMA showed that the evidence in favor of H1 was weak for any PDA (BF10 = 2.90; 10 studies), moderate for hemodynamically significant PDA (BF10 = 3.77; 3 studies), and extreme for surgically ligated or catheter-occluded PDA (BF10 = 294.9; 16 studies). In contrast, the evidence in favor of H0 was weak for medically treated PDA (BF10 = 0.55; 6 studies). In addition, BMA found strong evidence in favor of H1 when prolonged exposure to PDA was analyzed as a dichotomous variable (BF10 = 11.80; 6 studies) and extreme evidence (BF10 = 113.60; 3 studies) when PDA exposure time was analyzed as a continuous variable.Conclusions and RelevanceIn this bayesian meta-analysis, the data suggest that prolonged exposure to PDA might be associated with increased risk of pulmonary vascular disease in extremely preterm infants. This highlights the need to monitor for PH in high-risk preterm infants with prolonged exposure to PDA and to incorporate PH risk into clinical decisions regarding PDA management.

show abstract

A Primer on Bayesian Model-Averaged Meta-Analysis

Cited by 49 publications

References 72 publications

Benefits of Bayesian Model Averaging for Mixed-Effects Modeling

Benefits of Bayesian Model Averaging for Mixed-Effects Modeling

A Tutorial on Bayesian Model-Averaged Meta-Analysis in JASP

Patent Ductus Arteriosus and Bronchopulmonary Dysplasia–Associated Pulmonary Hypertension

Contact Info

Product

Resources

About