Estimating psychopathological networks: Be careful what you wish for

Abstract

Network models, in which psychopathological disorders are conceptualized as a complex interplay of psychological and biological components, have become increasingly popular in the recent psychopathological literature (Borsboom, et. al., 2011). These network models often contain significant numbers of unknown parameters, yet the sample sizes available in psychological research are limited. As such, general assumptions about the true network are introduced to reduce the number of free parameters. Incorporating these assumptions, however, means that the resulting network will lead to reflect the particular structure assumed by the estimation method—a crucial and often ignored aspect of psychopathological networks. For example, observing a sparse structure and simultaneously assuming a sparse structure does not imply that the true model is, in fact, sparse. To illustrate this point, we discuss recent literature and show the effect of the assumption of sparsity in three simulation studies.

Back to the Basics: Rethinking Partial Correlation Network Methodology

“…That is, the false positive rate (1 -speci city) of glasso is regularly around 20 -25%. Further, while Epskamp et al (2017) cautioned that assuming sparsity will result in false negatives if the true network is dense, our results suggest that levels of sparsity not typically seen in psychological applications (< 20% connectivity; Figure 2) are necessary for consistent model selection (although speci city declined slightly for the largest sample sizes). In the context of replication, high false positive rates (> 20%) obscure the ability to consistently replicate network structures.…”

“…When these covariances are standardized and the sign reversed, they correspond to partial correlations that imply pairwise dependencies in which the linear e ects of all other variables have been controlled for (Fan, Liao, & Liu, 2016;Whittaker, 1990). Since direct e ects allow for rich inferences, this has resulted in a growing body of literature called "network modeling" in both methodological (Epskamp & Fried, 2016;Epskamp, Kruis, & Marsman, 2017) and applied contexts (McNally et al, 2015;Rhemtulla et al, 2016).…”

Posttraumatic Stress Disorder Symptoms and Reckless Behaviors: A Network Analysis Approach

“…In several factor analytic studies, the hybrid model has outperformed other models, including the four‐factor DSM‐5 model (e.g., Contractor et al., 2018; Ito, Takebayashi, Suzuki, & Horikoshi, 2019), and several studies have shown the hybrid model's factors to have differential associations with external variables (e.g., Liu, Wang, Cao, Qing, & Armour, 2016; Zelazny & Simms, 2015) and substantial construct equivalence across compared subgroups (Contractor, Caldas, et al., 2019). Utilizing the seven‐factor hybrid model of PTSD rather than the 20 individual PTSD items reduces the number of parameters to be estimated in the network, which may increase the accuracy and stability of the estimated network when sample sizes are small (Epskamp, Borsboom, & Fried, 2018; Epskamp, Kruis, & Marsman, 2017). At the same time, compared with models that consist of fewer symptom clusters (e.g., the four‐factor DSM‐5 model; APA, 2013), the hybrid model, with its seven symptom clusters, provides an opportunity to examine the more fine‐grained associations between PTSD and reckless behaviors.…”

Replicability and generalizability of PTSD networks: A cross-cultural multisite study of PTSD symptoms in four trauma patient samples

“…Differences in sample size may also explain differences in network structures, because regularized partial correlation networks apply regularization procedures that act proportionately to power. When sample size goes to infinity, regularized and unregularized estimation procedures will result in very similar network structures, because even very small edges will be estimated reliably Epskamp, Kruis, & Marsman, 2016). In small samples, however, regularization will set even moderately large edge weights to zero, resulting in much sparser networks; this further complicates comparisons of network results across papers.…”