Learning to Live with Sampling Variability: Expected Replicability in Partial Correlation Networks

Dr, Williams

doi:10.31234/osf.io/fb4sa

Cited by 17 publications

(20 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is not surprising, as PTSD samples can consist of vastly different samples of subjects from different cultures exposed to different types of trauma. Second, this also means that, even when comparing high quality samples such that expected replicability given the estimation method is high (Williams, 2020), it cannot be expected PTSD network models replicate perfectly across such diverse samples. This contrasts previous discussion on PTSD network models (Forbes et al, 2017b), which discusses a need for perfectly reproducing network models across studies.…”

Section: Heterogeneity and Generalizability Of Ptsd Networkmentioning

confidence: 99%

Network Models of Post-traumatic Stress Disorder: A Meta-analysis

Isvoranu¹,

Epskamp²,

Cheung³

2020

Preprint

View full text Add to dashboard Cite

Post-traumatic 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 to the first-ever meta-analytic study of network models. 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 datasets, investigated the scope of between-study heterogeneity, and assessed the performance of network models estimated from single studies. Our main findings are that: (1) While several clear symptom-links and interpretable clusters can be identified in the network, most symptoms feature very similar levels of centrality. To this end, aiming to identify central symptoms in PTSD symptom networks may not be fruitful. (2) 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. (3) Nonetheless, we found evidence that networks estimated from single studies may give rise to generalizable results, as our results aligned with previous descriptive analyses of reported network studies, and network models estimated from single samples 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.

show abstract

Section: Heterogeneity and Generalizability Of Ptsd Networkmentioning

confidence: 99%

Network Models of Post-traumatic Stress Disorder: A Meta-analysis

Isvoranu¹,

Epskamp²,

Cheung³

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Here TP and FP denote the number of false positives, whereas TN and FN are the number of true and false negatives. For estimating A CD , note that on average specificity is equal to 1 − α (Figure 1 in and sensitivity is equal to the average power, 1 − β, across all edges (see Equation 7 in Williams, 2020). A CI was estimated by reversing the labels (0 and 1) in the true network.…”

Section: Model Selection Performancementioning

confidence: 99%

“…The topic of replicability has recently captivated the network literature (Forbes et al, 2019;Jones et al, 2019b;Williams, 2020). To assess replicability, it is common to focus on the individual edges with either classical (van Borkulo et al, 2016) or Bayesian testing (Williams et al, 2020).…”

Section: Replicating Edge Ordermentioning

confidence: 99%

Bayesian hypothesis testing for Gaussian graphical models: Conditional independence and order constraints

Williams

Mulder

2020

Journal of Mathematical Psychology

View full text Add to dashboard Cite

Gaussian graphical models (GGM; partial correlation networks) have become increasingly popular in the social and behavioral sciences for studying conditional (in)dependencies between variables. In this work, we introduce exploratory and confirmatory Bayesian tests for partial correlations. For the former, we first extend the customary GGM formulation that focuses on conditional dependence to also consider the null hypothesis of conditional independence for each partial correlation. Here a novel testing strategy is introduced that can provide evidence for a null, negative, or positive effect. We then introduce a test for hypotheses with order constraints on partial correlations. This allows for testing theoretical and clinical expectations in GGMs. The novel matrix−F prior distribution is described

show abstract

“…The nonconvex penalties depend on the initial estimates, which is not typically an issue in the most common situations, assuming it is "good enough" (p. 1511, Zou & Li, 2008). However, polychoric partials will typically have more sampling variance (e.g., Williams, 2020b), which suggests that the initial estimate for computing the derivative will also have more variance than Gaussian data. Here shrinkage estimators should be investigated (e.g., Ledoit & Wolf, 2004;Van Wieringen & Peeters, 2016).…”

Section: Future Directionsmentioning

confidence: 99%

Beyond Lasso: A Survey of Nonconvex Regularization in Gaussian Graphical Models

Dr¹

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Studying complex relations in multivariate datasets is a common task in psychological science. Recently, the Gaussian graphical model has emerged as an increasingly popular model for characterizing the conditional dependence structure of random variables. Although the graphical lasso ($\ell_1$-regularization) is the most well-known estimator across the sciences, it has several drawbacks that make it less than ideal for model selection. There are now alternative forms of regularization that were developed specifically to overcome issues inherent to the $\ell_1$-penalty.To date, this information has not been synthesized. This paper provides a comprehensive survey of nonconvex regularization that spans from the smoothly clipped absolute deviation penalty to continuous approximations of the $\ell_0$-penalty (i.e., best subset) for directly estimating the inverse covariance matrix. A common thread shared by these penalties is that they all enjoy the oracle properties, that is, they perform as though the \emph{true} generating model were known in advance. To ensure their theoretical properties are general, I conducted extensive numerical experiments that indicated superior and more than competitive performance when compared to glasso and non-regularized model selection, respectively, all the while being computationally feasible for many variables. In addition, the important topics of tuning parameter selection and statistical inference in regularized models are reviewed.The penalties are employed to estimate the dependence structure of post-traumatic stress disorder symptoms. The discussion includes several ideas for future research, including a plethora of information to facilitate their study. I have implemented the methods in the

show abstract

Learning to Live with Sampling Variability: Expected Replicability in Partial Correlation Networks

Cited by 17 publications

References 21 publications

Network Models of Post-traumatic Stress Disorder: A Meta-analysis

Network Models of Post-traumatic Stress Disorder: A Meta-analysis

Bayesian hypothesis testing for Gaussian graphical models: Conditional independence and order constraints

Beyond Lasso: A Survey of Nonconvex Regularization in Gaussian Graphical Models

Contact Info

Product

Resources

About