Objective Bayesian Edge Screening and Structure Selection for Ising Networks

Marsman, Maarten; Huth, Karoline; Waldorp, Lourens J.; Ntzoufras, Ioannis

doi:10.1007/s11336-022-09848-8

Cited by 19 publications

(17 citation statements)

References 84 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The term on the right is the posterior odds, which indicates the relative plausibility of the rival models after having seen the data. In this paper, we assume that the prior odds are equal to one by assuming p(S s ) = p(S t ), which makes the Bayes factor equal to the posterior odds (see Marsman, Huth, Waldorp, & Ntzoufras, 2022, for a different approach). The subscripts in the Bayes factor notation indicate in which direction the support is expressed.…”

Section: Bayesian Hypothesis Testing: the Bayes Factormentioning

confidence: 99%

Testing Conditional Independence in Psychometric Networks: An Analysis of Three Bayesian Methods

Sekulovski¹,

Keetelaar²,

Huth³

et al. 2023

Preprint

View full text Add to dashboard Cite

Network psychometrics uses graphical models to assess the network structure of psychological variables. An important task in their analysis is determining which variables are unrelated in the network, i.e., are independent given the rest of the network variables. This conditional independence structure is a gateway to understanding the causal structure underlying psychological processes. Thus, it is crucial to have an appropriate method for evaluating conditional independence and dependence hypotheses. Bayesian approaches to testing such hypotheses allow researchers to differentiate between absence of evidence and evidence of absence of connections (edges) between pairs of variables in a network. Three Bayesian approaches to assessing conditional independence have been proposed in the network psychometrics literature.We believe that their theoretical foundations are not widely known, and therefore we provide a conceptual review of the proposed methods and highlight their strengths and limitations through a small-scale simulation study. Finally, we provide recommendations on how to choose the optimal method and discuss the current gaps in the literature on this important topic.

show abstract

Section: Bayesian Hypothesis Testing: the Bayes Factormentioning

confidence: 99%

Testing Conditional Independence in Psychometric Networks: An Analysis of Three Bayesian Methods

Sekulovski¹,

Keetelaar²,

Huth³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Several papers in the special issue aim to fill that gap. For example, Marsman et al (2022; this issue) offer Bayesian solutions for assessing edge inclusion for the Ising model, a network for binary variables, addressing similar questions as Williams and Mulder (2020a) did for GGMs. While Epskamp et al (2022, this issue) offer a classical approach to gauge the heterogeneity of a GGM applied to independent datasets, Lee, Chen, DeSarbo, and Xue (2022; this issue) gauge the heterogeneity of networks of ordinal variables estimated from cross-sectional data.…”

Section: How Can We Conduct Confirmatory Tests Of the Relationship Be...mentioning

confidence: 99%

Guest Editors’ Introduction to The Special Issue “Network Psychometrics in Action”: Methodological Innovations Inspired by Empirical Problems

Marsman

Rhemtulla

2022

Psychometrika

View full text Add to dashboard Cite

In 1904, Spearman reported the positive manifold-a pattern of exclusively positive correlations among cognitive test scores-for which he proposed the g-factor theory of general intelligence in 1927. According to g-factor theory, general intelligence is a stable trait that cannot be directly observed but gives rise to observed test scores. The g-factor model quickly became the dominant theory of intelligence and, while the theory was revised and expanded, the fundamental idea went unchallenged for almost a century. Modifications to the g-factor theory (e.g., multifactor models, hierarchical factor models, and their merger) were proposed and debated at length (e.g.,

show abstract

“…Altogether this paper pointed to the necessity of solid uncertainty quantification in complex statistical models, such as networks, to not draw unwarranted inferential conclusions. Bayesian approaches offer an avenue to address this issue by allowing to quantify structure uncertainty, determine parameter stability, obtain edge inclusion evidence and acknowledging structure uncertainty in further inferential decision through Bayesian model-averaging (Marsman et al, 2022;Williams et al, 2021).…”

Section: Limitationsmentioning

confidence: 99%

Sociodemographic Factors and Alcohol Use Symptoms in the HELIUS study: A Bayesian Analysis of Networks

Huth¹,

Zavlis²,

Luigjes³

et al. 2022

Preprint

View full text Add to dashboard Cite

Problematic alcohol use is argued to develop according to complex system principles, where clinical phenomena emerge through a bidirectional, (non-)linear interplay between psychological symptoms, (neuro-)biological factors, and environmental components including sociodemographic determinants, for example, ethnic, religious, and socioeconomic variables. These principles can be modelled through the network perspective. Research remains uncertain how ethnic, religious, and socioeconomic variables influence and are influenced by symptoms of alcohol use disorder, and how these interactions vary between ethnic groups. To investigate these open-standing questions, we conduct an exploratory network analysis in a Bayesian framework in a large, multi-ethnic, urban sample in the Netherlands, the HELIUS study (N = 22,164). Our analyses revealed that sociodemographic factors were differentially related to alcohol use symptoms, with religion being negatively associated to binge drinking and adverse events due to drinking (e.g., feelings of guilt). Sex differences were revealed: higher education buffered against more adverse AUD symptoms for females and being employed for males. Stratifying our analyses by ethnicity, results suggested a different importance of sociodemographic determinants for alcohol use symptoms across ethnic groups; whereas for some subgroups all factors were important others showed hardly any interactions. Considering the trend of urban areas becoming increasingly diverse, our results point to the finding that sociodemographic factors are important to acknowledge for particular alcohol use symptoms. Tailored sex-specific prevention and intervention strategies should be considered and evaluated for urban ethnic minority communities.

show abstract

Objective Bayesian Edge Screening and Structure Selection for Ising Networks

Cited by 19 publications

References 84 publications

Testing Conditional Independence in Psychometric Networks: An Analysis of Three Bayesian Methods

Testing Conditional Independence in Psychometric Networks: An Analysis of Three Bayesian Methods

Guest Editors’ Introduction to The Special Issue “Network Psychometrics in Action”: Methodological Innovations Inspired by Empirical Problems

Sociodemographic Factors and Alcohol Use Symptoms in the HELIUS study: A Bayesian Analysis of Networks

Contact Info

Product

Resources

About