Estimating psychopathological networks: Be careful what you wish for

A network psychometric approach to neurocognition in early Alzheimer's disease

In a typical pattern of Alzheimer's disease onset, episodic memory decline is predominant while decline in other neurocognitive domains is subsidiary or absent. Such descriptions refer to relationships between neurocognitive domains as well as deficits within domains. However, the former relationships are rarely statistically modelled. This study used psychometric network analysis to model relationships between neurocognitive variables in cognitive normality (CN), amnestic mild cognitive impairment (aMCI), and early Alzheimer's disease (eAD). Gaussian graphical models with extended Bayesian information criterion graphical lasso model selection and regularisation were used to estimate network models of neurocognitive and demographic variables in CN (n = 229), aMCI (n = 395), and eAD (n = 191) groups. The edge density, network strength and structure, centrality, and individual links of the network models were explored. Results indicated that while global strength did not differ, network structures differed across CN and eAD and aMCI and eAD groups, suggesting neurocognitive reorganisation across the eAD continuum. Episodic memory variables were most central (i.e., influential) in the aMCI network model, whereas processing speed and fluency variables were most central in the eAD network model. Additionally, putative clusters of memory, language and semantic variables, and attention, processing speed and working memory variables arose in the models for the clinical groups. This exploratory study shows how psychometric network analysis can be used to model the relationships between neurocognitive variables across the eAD continuum and to generate hypotheses for future (dis)confirmatory research.

On the equivalency of factor and network loadings

Recent research has demonstrated that the network measure node strength or sum of a node's connections is roughly equivalent to confirmatory factor analysis (CFA) loadings. A key finding of this research is that node strength represents a combination of different latent causes. In the present research, we sought to circumvent this issue by formulating a network equivalent of factor loadings, which we call network loadings. In two simulations, we evaluated whether these network loadings could effectively (1) separate the effects of multiple latent causes and (2) estimate the simulated factor loading matrix of factor models. Our findings suggest that the network loadings can effectively do both. In addition, we leveraged the second simulation to derive effect size guidelines for network loadings. In a third simulation, we evaluated the similarities and differences between factor and network loadings when the data were generated from random, factor, and network models. We found sufficient differences between the loadings, which allowed us to develop an algorithm to predict the data generating model called the Loadings Comparison Test (LCT). The LCT had high sensitivity and specificity when predicting the data generating model. In sum, our results suggest that network loadings can provide similar information to factor loadings when the data are generated from a factor model and therefore can be used in a similar way (e.g., item selection, measurement invariance, factor scores).

Psychological networks in clinical populations: investigating the consequences of Berkson's bias

BACKGROUND

In clinical research, populations are often selected on the sum-score of diagnostic criteria such as symptoms. Estimating statistical models where a subset of the data is selected based on a function of the analyzed variables introduces Berkson's bias, which presents a potential threat to the validity of findings in the clinical literature. The aim of the present paper is to investigate the effect of Berkson's bias on the performance of the two most commonly used psychological network models: the Gaussian Graphical Model (GGM) for continuous and ordinal data, and the Ising Model for binary data.

METHODS

In two simulation studies, we test how well the two models recover a true network structure when estimation is based on a subset of the data typically seen in clinical studies. The network is based on a dataset of 2807 patients diagnosed with major depression, and nodes in the network are items from the Hamilton Rating Scale for Depression (HRSD). The simulation studies test different scenarios by varying (1) sample size and (2) the cut-off value of the sum-score which governs the selection of participants.

RESULTS

The results of both studies indicate that higher cut-off values are associated with worse recovery of the network structure. As expected from the Berkson's bias literature, selection reduced recovery rates by inducing negative connections between the items.

CONCLUSION

Our findings provide evidence that Berkson's bias is a considerable and underappreciated problem in the clinical network literature. Furthermore, we discuss potential solutions to circumvent Berkson's bias and their pitfalls.

Identifying Fundamental Motor Skills Building Blocks in Preschool Children From Brazil and the United States: A Network Analysis

Fundamental motor skills (FMSs) are building blocks for future movements and may vary according to cultural context. Moreover, network analysis can identify which skills contribute most to an overall set of skills. This study identified the most influential FMS in samples of U.S. and Brazil preschoolers that may contribute to a pattern of adequate motor skills. Participants were 101 Brazilian (55 boys; 47.52 ± 5.57 months of age) and 236 U.S. preschoolers (108 boys; 49.56 ± 8.27 months of age), who provided completed FMS assessments (Test of Gross Motor Development—third edition). Confirmatory factorial analysis was used to test alternative models. To quantify the importance of each variable in the network, the expected influence was calculated, using the network analysis Mplus, Rstudio, and JASP (version 0.14.1). Reduced models with nine and 11 FMS for Brazilian and U.S. preschoolers, respectively, showed adequate adjustment indexes. Jump (1.412) and one-hand strike (0.982) in the Brazilian sample, and hop (1.927) and dribble (0.858) in the U.S. sample, showed the highest expected influence values. This study presents a new perspective to report which are the most important FMS in preschoolers of different sociocultural contexts, which act as building blocks for the acquisition of more complex motor skills.

The Association Between Internet Addiction and Anxiety in Nursing Students: A Network Analysis

Background: Nursing students who suffer from co-occurring anxiety experience added difficulties when communicating and interacting with others in a healthy, positive, and meaningful way. Previous studies have found strong positive correlations between Internet addiction (IA) and anxiety, suggesting that nursing students who report severe IA are susceptible to debilitating anxiety as well. To date, however, network analysis (NA) studies exploring the nature of association between individual symptoms of IA and anxiety have not been published. Objective: This study examined associations between symptoms of IA and anxiety among nursing students using network analysis. Methods: IA and anxiety symptoms were assessed using the Internet Addiction Test (IAT) and the Generalized Anxiety Disorder Screener (GAD-7), respectively. The structure of IA and anxiety symptoms was characterized using "Strength" as a centrality index in the symptom network. Network stability was tested using a case-dropping bootstrap procedure and a Network Comparison Test (NCT) was conducted to examine whether network characteristics differed on the basis of gender and by region of residence. Results: A total of 1,070 nursing students participated in the study. Network analysis showed that IAT nodes, "Academic decline due to Internet use," "Depressed/moody/nervous only while being off-line," "School grades suffer due to Internet use," and "Others complain about your time spent online" were the most influential symptoms in the IA-anxiety network model. Gender and urban/rural residence did not significantly influence the overall network structure. Conclusion: Several influential individual symptoms including Academic declines due to Internet use, Depressed/moody/nervous only while being off-line, School grades suffering due to Internet use and Others complain about one's time spent online emerged as potential targets for clinical interventions to reduce co-occurring IA and anxiety. Additionally, the overall network structure provides a data-based hypothesis for explaining potential mechanisms that account for comorbid IA and anxiety.

Pediatric health-related quality of life and school social capital through network perspectives

Background

Despite their importance in population health among children and adolescents, our understanding of how individual items mutually interact within and between pediatric health-related quality of life (HRQOL) and school social capital is limited.

Methods

We employed network analysis in a general population sample of 7759 children aged 9–15 years to explore the network structure of relations among pediatric HRQOL and school social capital items measured using validated scales. Furthermore, network centrality was examined to identify central items that had stronger and more direct connections with other items in the network than others. Network structure and overall strength of connectivity among items were compared between groups (by sex and age).

Results

Our analysis revealed that the item related to school/academic functioning and the item related to shared enjoyment among students had the highest strength centrality in the network of HRQOL and school social capital, respectively, underpinning their critical roles in pediatric HRQOL and school social capital. Additionally, the edge connecting “I trust my friends at school” and “trouble getting along with peers” had the strongest negative edge weight among ones connecting school social capital and pediatric HRQOL constructs. Network comparison test revealed stronger overall network connectivity in middle schoolers compared to elementary schoolers but no differences between male and female students.

Conclusion

The network approach elucidated the complex relationship of mutually influencing items within and between pediatric HRQOL and school social capital. Addressing central items may promote children’s perceived health and school social capital.

Resilience and Stress in Later Life: A Network Analysis Approach Depicting Complex Interactions of Resilience Resources and Stress-Related Risk Factors in Older Adults

BACKGROUND

Emerging systemic approaches on resilience propose that a person's or group's adaptability to significant stress relies on a network of interdependent resources. However, little knowledge exists on systemic resilience in older survivors of early-life adversity (ELA) and how ELA affects their resource network in later life.

OBJECTIVE

This study investigated how ELA may be linked to the interplay of resources and stress-related risk factors in later life.

RESEARCH DESIGN AND METHODS

Data from N = 235 older adults (M age = 70.43 years; 46.40% female) were assessed. Half the participants were affected by ELA through compulsory social measures and placements in childhood, and/or adolescence ("risk group"). The other half were age-matched, non-affected participants ("control group"). Using psychometric instruments, a set of resilience-supporting resources in later life and current stress indices were assessed. Regularized partial correlation networks examined the interplay of resources in both groups, whilst also considering the impact of stress.

RESULTS

Both groups demonstrated only positive resource interrelations. Although the control group showed more possible resource connections, the groups did not significantly differ in the overall strength of connections. While group-specific resource interrelations were identified, self-esteem was observed to be the most important resource for the network interconnectedness of both groups. The risk group network showed a higher vulnerability to current stress.

DISCUSSION AND IMPLICATIONS

Network analysis is a useful approach in the examination of the complex interrelationships between resilience resources and stress-related risk factors in older adulthood.

Network Analysis of ADHD and ODD Symptoms: Novel Insights or Redundant Findings with the Latent Variable Model?

A latent variable model (LVM) and network analysis (NA) were applied to mother and father ratings of attention-deficit/hyperactivity disorder (ADHD) and oppositional defiant disorder (ODD) symptoms to determine if NA offers unique insights relative to the LVM. ADHD-inattention (IN), ADHD-hyperactivity/impulsivity (HI), and ODD symptoms along with academic competence behaviors (reading, arithmetic, and writing skills) were rated by mothers and fathers of Brazilian (n = 894), Thai (n = 2075), and United States (n = 817) children (M_age = 9.04, SD = 2.12, 49.5% females). LVM indicated that (1) the ADHD-IN, ADHD-HI, and ODD three-factor model yielded a close global-fit with no localized ill-fit; (2) nearly all loadings were substantial; (3) like-symptom loadings, like-symptom thresholds, and like-factor means showed invariance across mothers and fathers; (4) the three factors showed convergent and discriminant validity across mothers and fathers; and (5) only the ADHD-IN showed a unique negative relationship with academic competence. NA indicated that (1) a walktrap community analysis resulted in ADHD-IN, ADHD-HI, and ODD symptom communities; (2) the three symptom communities were consistent across mothers and fathers; (3) only three ADHD-IN symptoms showed unique relationships with the three academic competence items. NA has proven useful for numerous mental disorders. In the current study, NA results were mostly congruent with the LVM model, with a few notable exceptions. The results are discussed in the context of model assumptions and application considerations in the context of ADHD/ODD symptoms relative to other symptom dimensions.

Network Analysis of Internet Addiction Symptoms Among a Clinical Sample of Japanese Adolescents with Autism Spectrum Disorder

In the present study, we employed network analysis that conceptualizes internet addiction (IA) as a complex network of mutually influencing symptoms in 108 adolescents with autism spectrum disorder (ASD) to examine the network architecture of IA symptoms and identify central/influential symptoms. Our analysis revealed that defensive and secretive behaviors and concealment of internet use were identified as central symptoms in this population, suggesting that mitigating these symptoms potentially prevent the development and/or maintenance of IA in adolescents with ASD. Providing adolescents and their caregivers with psychoeducation on the role of central symptoms above in IA can be a salient intervention. Doing so may facilitate nonconflicting conversations between them about adolescents' internet use and promote more healthy adolescents' internet use behavior.

A network model of affective odor perception

The affective appraisal of odors is known to depend on their intensity (I), familiarity (F), detection threshold (T), and on the baseline affective state of the observer. However, the exact nature of these relations is still largely unknown. We therefore performed an observer experiment in which participants (N = 52) smelled 40 different odors (varying widely in hedonic valence) and reported the intensity, familiarity and their affective appraisal (valence and arousal: V and A) for each odor. Also, we measured the baseline affective state (valence and arousal: BV and BA) and odor detection threshold of the participants. Analyzing the results for pleasant and unpleasant odors separately, we obtained two models through network analysis. Several relations that have previously been reported in the literature also emerge in both models (the relations between F and I, F and V, I and A; I and V, BV and T). However, there are also relations that do not emerge (between BA and V, BV and I, and T and I) or that appear with a different polarity (the relation between F and A for pleasant odors). Intensity (I) has the largest impact on the affective appraisal of unpleasant odors, while F significantly contributes to the appraisal of pleasant odors. T is only affected by BV and has no effect on other variables. This study is a first step towards an integral study of the affective appraisal of odors through network analysis. Future studies should also include other factors that are known to influence odor appraisal, such as age, gender, personality, and culture.

The symptom network structure of depressive symptoms in late-life: Results from a European population study

The network theory conceptualizes mental disorders as complex networks of symptoms influencing each other by creating feedback loops, leading to a self-sustained syndromic constellation. Symptoms central to the network have the greatest impact in sustaining the rest of symptoms. This analysis focused on the network structure of depressive symptoms in late-life because of their distinct etiologic factors, clinical presentation, and outcomes. We analyzed cross-sectional data from wave 2 of the 19 country Survey of Health, Ageing, and Retirement in Europe (SHARE) and included non-institutionalized adults aged 65 years or older (mean age 74 years, 59% females) endorsing at least one depressive symptom on the EURO-D scale for depression (N =8,557). We characterized the network structure of depressive symptoms in late-life and used indices of "strength", "betweenness", and "closeness" to identify symptoms central to the network. We used a case-dropping bootstrap procedure to assess network stability. Death wishes, depressed mood, loss of interest, and pessimism had the highest values of centrality. Insomnia, fatigue and appetite changes had lower centrality values. The identified network remained stable after dropping 74.5% of the sample. Sex or age did not significantly influence the network structure. In conclusion, death wishes, depressed mood, loss of interest, and pessimism constitute the "backbone" that sustains depressive symptoms in late-life. Symptoms central to the network of depressive symptoms may be used as targets for novel, focused interventions and in studies investigating neurobiological processes central to late-life depression.

Can network analysis shed light on predictors of lithium response in bipolar I disorder?

OBJECTIVE

To undertake a large-scale clinical study of predictors of lithium (Li) response in bipolar I disorder (BD-I) and apply contemporary multivariate approaches to account for inter-relationships between putative predictors.

METHODS

We used network analysis to estimate the number and strength of connections between potential predictors of good Li response (measured by a new scoring algorithm for the Retrospective Assessment of Response to Lithium Scale) in 900 individuals with BD-I recruited to the Consortium of Lithium Genetics.

RESULTS

After accounting for co-associations between potential predictors, the most important factors associated with the good Li response phenotype were panic disorder, manic predominant polarity, manic first episode, age at onset between 15-32 years and family history of BD. Factors most strongly linked to poor outcome were comorbid obsessive-compulsive disorder, alcohol and/or substance misuse, and/or psychosis (symptoms or syndromes).

CONCLUSIONS

Network analysis can offer important additional insights to prospective studies of predictors of Li treatment outcomes. It appears to especially help in further clarifying the role of family history of BD (i.e. its direct and indirect associations) and highlighting the positive and negative associations of different subtypes of anxiety disorders with Li response, particularly the little-known negative association between Li response and obsessive-compulsive disorder.

The Network Structure of Personality Pathology in Adolescence With the 100-Item Personality Inventory for DSM-5 Short-Form (PID-5-SF)

There is currently a lack of understanding of the structure of personality disorder (PD) trait facets. The network approach may be useful in providing additional insights, uncovering the unique association of each PD trait facet with every other facet. A unique feature of network analysis is centrality, which indicates the importance of the role a trait facet plays in the context of other trait facets. Using data from 1,940 community Dutch adolescents, we applied network analysis to the 25 trait facets from the 100-item Personality Inventory for DSM-5 Short-Form (PID-5-SF) to explore their associations. We found that some trait facets only seem to be core indicators of their pre-ordained domains, whereas we observed that other trait facets were strongly associated with trait facets outside of their hypothesized domains. Importantly, anxiousness and callousness were identified as highly central facets, being uniquely associated with many other trait facets. Future longitudinal network studies could therefore further examine the possibility of anxiousness and callousness as risk marker trait facets among other PD trait facets.

Back to the basics: Rethinking partial correlation network methodology

The Gaussian graphical model (GGM) is an increasingly popular technique used in psychology to characterize relationships among observed variables. These relationships are represented as elements in the precision matrix. Standardizing the precision matrix and reversing the sign yields corresponding partial correlations that imply pairwise dependencies in which the effects of all other variables have been controlled for. The graphical lasso (glasso) has emerged as the default estimation method, which uses ℓ1 -based regularization. The glasso was developed and optimized for high-dimensional settings where the number of variables (p) exceeds the number of observations (n), which is uncommon in psychological applications. Here we propose to go 'back to the basics', wherein the precision matrix is first estimated with non-regularized maximum likelihood and then Fisher Z transformed confidence intervals are used to determine non-zero relationships. We first show the exact correspondence between the confidence level and specificity, which is due to 1 minus specificity denoting the false positive rate (i.e., α). With simulations in low-dimensional settings (p ≪ n), we then demonstrate superior performance compared to the glasso for detecting the non-zero effects. Further, our results indicate that the glasso is inconsistent for the purpose of model selection and does not control the false discovery rate, whereas the proposed method converges on the true model and directly controls error rates. We end by discussing implications for estimating GGMs in psychology.

Eating disorder psychopathology in adults and adolescents with anorexia nervosa: A network approach

OBJECTIVE

The aim of this study was to assess and compare eating disorder feature networks in adult and adolescent patients with anorexia nervosa.

METHODS

Patients seeking treatment for anorexia nervosa in inpatient and outpatient settings were consecutively recruited from January 2008 to September 2019. Body mass index was measured, and each patient completed the Eating Disorder Examination Questionnaire.

RESULTS

The sample comprised 547 adolescent and 724 adult patients with anorexia nervosa. Network analysis showed that in both adults and adolescents, the most central and highly interconnected nodes in the network were related to shape overvaluation and desiring weight loss. The network comparison test identified similar global strength and network invariance, confirming the similarity of the two network structures.

DISCUSSION

The network structures in adult and adolescent patients with anorexia nervosa are similar, and lend weight to the cognitive behavioral theory that overvaluation of shape and weight is the core feature of anorexia nervosa psychopathology.

All Happy Emotions Are Alike but Every Unhappy Emotion Is Unhappy in Its Own Way: A Network Perspective to Academic Emotions

Quantitative research into the nature of academic emotions has thus far been dominated by factor analyses of questionnaire data. Recently, psychometric network analysis has arisen as an alternative method of conceptualizing the composition of psychological phenomena such as emotions: while factor models view emotions as underlying causes of affects, cognitions and behavior, in network models psychological phenomena are viewed as arising from the interactions of their component parts. We argue that the network perspective is of interest to studies of academic emotions due to its compatibility with the theoretical assumptions of the control value theory of academic emotions. In this contribution we assess the structure of a Finnish questionnaire of academic emotions using both network analysis and exploratory factor analysis on cross-sectional data obtained during a single course. The global correlational structure of the network, investigated using the spinglass community detection analysis, differed from the results of the factor analysis mainly in that positive emotions were grouped in one community but loaded on different factors. Local associations between pairs of variables in the network model may arise due to different reasons, such as variable A causing variation in variable B or vice versa, or due to a latent variable affecting both. We view the relationship between feelings of self-efficacy and the other emotions as causal hypotheses, and argue that strengthening the students' self-efficacy may have a beneficial effect on the rest of the emotions they experienced on the course. Other local associations in the network model are argued to arise due to unmodeled latent variables. Future psychometric studies may benefit from combining network models and factor models in researching the structure of academic emotions.

Personalized models of psychopathology as contextualized dynamic processes: An example from individuals with borderline personality disorder

OBJECTIVE

Psychopathology research has relied on discrete diagnoses, which neglects the unique manifestations of each individual's pathology. Borderline personality disorder combines interpersonal, affective, and behavioral regulation impairments making it particularly ill-suited to a "one size fits all" diagnosis. Clinical assessment and case formulation involve understanding and developing a personalized model for each patient's contextualized dynamic processes, and research would benefit from a similar focus on the individual.

METHOD

We use group iterative multiple model estimation, which estimates a model for each individual and identifies general or shared features across individuals, in both a mixed-diagnosis sample (N = 78) and a subsample with a single diagnosis (n = 24).

RESULTS

We found that individuals vary widely in their dynamic processes in affective and interpersonal domains both within and across diagnoses. However, there was some evidence that dynamic patterns relate to transdiagnostic baseline measures. We conclude with descriptions of 2 person-specific models as an example of the heterogeneity of dynamic processes.

CONCLUSIONS

The idiographic models presented here join a growing literature showing that the individuals differ dramatically in the total patterning of these processes, even as key processes are shared across individuals. We argue that these processes are best estimated in the context of person-specific models, and that so doing may advance our understanding of the contextualized dynamic processes that could identify maintenance mechanisms and treatment targets. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Psychological features in obesity: A network analysis

OBJECTIVE

Enhanced understanding of psychosocial factors associated with obesity may improve knowledge of their interplay mechanisms. The aim of this study was to assess the relationship between psychosocial variables in individuals with obesity using a network analysis.

METHODS

Patients seeking treatment for obesity were consecutively recruited from a rehabilitative residential treatment program for severe obesity between January 2016 and March 2019. Each patient completed the following questionnaires: Eating Disorder Examination Questionnaire, Symptom Checklist-90, Obesity Related Well-Being, and Weight Bias Internalization Scale. In addition, current body mass index (BMI) was measured, and maximum acceptable and dream BMI were assessed.

RESULTS

The sample comprised 996 patients with obesity (age 52.3 [SD = 16.0] years; BMI 41.8 [SD = 7.8] kg/m² ; 65.7% women; 52.2% married or living with a partner). Network analysis showed that interpersonal sensitivity and shape-weight concern, but also internalized weight stigma, were the most central and highly interconnected nodes in the network. In contrast, objective binge-eating episodes and dietary restraint were the most peripheral and least connected nodes. Eating disorder features and psychological distress formed two clearly separate clusters. No difference in network structure was found between men and women.

CONCLUSIONS

The pattern of network node connections supports the importance of assessing psychological distress, interpersonal sensitivity, shape-weight concern, and internalized weight stigma in patients seeking treatment for obesity.

Examining the boundaries between ICD-11 PTSD/CPTSD and depression and anxiety symptoms: A network analysis perspective

BACKGROUND

Two newly identified sibling disorders - ICD-11 PTSD and CPTSD - have been well validated in the last few years. Although these trauma-related disorders are suggested to be neatly separated from depression and anxiety, no study has used a network analysis to examine those definitions' construct validity when they also interplay with symptoms of depression and anxiety. Additionally, no research has focused upon the specific boundaries between these four disorders' symptoms, the bridges between them, and the ways they influence each other among clinical populations.

METHODS

A sample of 234 men drawn randomly from a national sample of 1,600 Jewish men receiving treatment for domestic violence in Israel completed the ICD-11 International Trauma Questionnaire (ITQ) and Brief Symptom Inventory (BSI).

RESULTS

The ICD-11 CPTSD, depression and anxiety clustering network results revealed, within the EGA, a four-cluster solution in which PTSD and CPTSD symptoms are differentiated from two other distinct clusters of anxiety and depression symptoms. Feelings of worthlessness and avoiding internal reminders of the experience were the most central symptoms.

LIMITATIONS

Due to the use of a cross-sectional design, causal interpretation of the network correlation between symptoms should be made cautiously.

CONCLUSIONS

These findings strengthen the approach that ICD-11 PTSD and CPTSD have a distinct construct; however, they also reflect a strong positive connection to anxiety and depression symptoms and no clear boundaries between disorders. Specifically, dysphoria/avoidance-related symptoms act as a bridge between the disorders, which may be important targets for specific assessments and related interventions.

The network approach to psychopathology: a review of the literature 2008-2018 and an agenda for future research

The network approach to psychopathology posits that mental disorders can be conceptualized and studied as causal systems of mutually reinforcing symptoms. This approach, first posited in 2008, has grown substantially over the past decade and is now a full-fledged area of psychiatric research. In this article, we provide an overview and critical analysis of 363 articles produced in the first decade of this research program, with a focus on key theoretical, methodological, and empirical contributions. In addition, we turn our attention to the next decade of the network approach and propose critical avenues for future research in each of these domains. We argue that this program of research will be best served by working toward two overarching aims: (a) the identification of robust empirical phenomena and (b) the development of formal theories that can explain those phenomena. We recommend specific steps forward within this broad framework and argue that these steps are necessary if the network approach is to develop into a progressive program of research capable of producing a cumulative body of knowledge about how specific mental disorders operate as causal systems.

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

Existing literature indicates a theoretical and empirical relation between engagement in reckless behaviors and posttraumatic stress disorder (PTSD). Thus, the DSM-5 revision of the PTSD nosology added a new "reckless or self-destructive behavior" (RSDB) symptom (Criterion E2). The current study applied a network analytic approach to examine the item-level relations among a range of reckless behaviors and PTSD symptom clusters. Participants were recruited from Amazon's Mechanical Turk (N = 417), and network analysis was conducted with 20 variables: six PTSD symptom clusters, corresponding to the hybrid model of PTSD (Armour et al., 2015) and excluding the externalizing behavior cluster (Community 1), and 14 items related to reckless behavior (Community 2). The results showed that the network associations were strongest within each construct (i.e., within PTSD and within reckless behaviors), although several bridge connections (i.e., between PTSD clusters and reckless behaviors) were identified. Most reckless behavior items had direct associations with one or more PTSD symptom clusters. The present findings support the existence of close relations between a variety of reckless behaviors and PTSD symptom clusters beyond their relations with DSM Criterion E2. The results provide testable hypotheses about the associations between specific reckless behaviors and PTSD symptom clusters, which may inform future research.

The replicability and generalizability of internalizing symptom networks across five samples

The popularity of network analysis in psychopathology research has increased exponentially in recent years. Yet, little research has examined the replicability of cross-sectional psychopathology network models, and those that have used single items for symptoms rather than multiitem scales. The present study therefore examined the replicability and generalizability of regularized partial correlation networks of internalizing symptoms within and across 5 samples (total N = 2,573) using the Inventory for Depression and Anxiety Symptoms, a factor analytically derived measure of individual internalizing symptoms. As different metrics may yield different conclusions about the replicability of network parameters, we examined both global and specific metrics of similarity between networks. Correlations within and between nonclinical samples suggested considerable global similarities in network structure (rss = .53-.87) and centrality strength (rss = .37-.86), but weaker similarities in network structure (rss = .36-.66) and centrality (rss = .04-.54) between clinical and nonclinical samples. Global strength (i.e., connectivity) did not significantly differ across all 5 networks and few edges (0-5.5%) significantly differed between networks. Specific metrics of similarity indicated that, on average, approximately 80% of edges were consistently estimated within and between all 5 samples. The most central symptom (i.e., dysphoria) was consistent within and across samples, but there were few other matches in centrality rank-order. In sum, there were considerable similarities in network structure, the presence and sign of individual edges, and the most central symptom within and across internalizing symptom networks estimated from nonclinical samples, but global metrics suggested network structure and symptom centrality had weak to moderate generalizability from nonclinical to clinical samples. (PsycINFO Database Record (c) 2020 APA, all rights reserved).

Application of network analysis to investigate sex differences in interactive systems of eating-disorder psychopathology

OBJECTIVE

Although men comprise 25% of persons with eating disorders (EDs), most research has focused on understanding EDs in women. The theoretical framework underlying common ED treatment has not been rigorously tested in men. The purpose of this study was to compare the interconnectivity among ED symptoms in men versus women.

METHOD

Participants (N = 1,348; 50% men) were individuals with anorexia nervosa, bulimia nervosa, binge-eating disorder, or other specified feeding or eating disorder who were users of Recovery Record, a smartphone app for monitoring ED symptoms. Participants were matched on age and duration of illness. Network analysis was used to create networks of symptoms for both sexes. Strength centrality, network stability, and bootstrapped centrality differences were tested. The network comparison test (NCT) was used to identify sex differences between networks. Key players analysis was used to compare fragmentation of each network.

RESULTS

For both sexes, items related to binge eating and restricting emerged as highest in strength centrality. The NCT identified significant differences global strength (p = .03) but not network invariance (p = .06) suggesting that although the structure of the networks was not statistically different, the strength of the connections within the network was greater for women. Key players analysis indicated that both networks were similarly disrupted when important nodes within the network were removed.

DISCUSSION

Findings suggested that there are more similarities than differences in networks of EDs in men and women. Results have important clinical implications by supporting theoretical underpinnings of cognitive-behavioral models of EDs in both men and women.

A Partial Correlation Screening Approach for Controlling the False Positive Rate in Sparse Gaussian Graphical Models

Gaussian Graphical Models (GGMs) are extensively used in many research areas, such as genomics, proteomics, neuroimaging, and psychology, to study the partial correlation structure of a set of variables. This structure is visualized by drawing an undirected network, in which the variables constitute the nodes and the partial correlations the edges. In many applications, it makes sense to impose sparsity (i.e., some of the partial correlations are forced to zero) as sparsity is theoretically meaningful and/or because it improves the predictive accuracy of the fitted model. However, as we will show by means of extensive simulations, state-of-the-art estimation approaches for imposing sparsity on GGMs, such as the Graphical lasso, ℓ1 regularized nodewise regression, and joint sparse regression, fall short because they often yield too many false positives (i.e., partial correlations that are not properly set to zero). In this paper we present a new estimation approach that allows to control the false positive rate better. Our approach consists of two steps: First, we estimate an undirected network using one of the three state-of-the-art estimation approaches. Second, we try to detect the false positives, by flagging the partial correlations that are smaller in absolute value than a given threshold, which is determined through cross-validation; the flagged correlations are set to zero. Applying this new approach to the same simulated data, shows that it indeed performs better. We also illustrate our approach by using it to estimate (1) a gene regulatory network for breast cancer data, (2) a symptom network of patients with a diagnosis within the nonaffective psychotic spectrum and (3) a symptom network of patients with PTSD.

A Clinician’s Primer for Idiographic Research: Considerations and Recommendations

Theorists and clinicians have long noted the need for idiographic (i.e., individual-level) designs within clinical psychology. Results from idiographic work may provide a possible resolution of the therapist's dilemma-the problem of treating an individual using information gathered via group-level research. Due to advances in data collection and time series methodology, there has been increasing interest in using idiographic designs to answer clinical questions. Although time series methods have been well-studied outside the field of clinical psychology, there is limited direction on how clinicians can use such models to inform their clinical practice. In this primer, we collate decades of published and word-of-mouth information on idiographic designs, measurement, and modeling. We aim to provide an initial guide on the theoretical and practical considerations that we urge interested clinicians to consider before conducting idiographic work of their own.

Characterizing the Manifest Probability Distributions of Three Latent Trait Models for Accuracy and Response Time

In this paper we study the statistical relations between three latent trait models for accuracies and response times: the hierarchical model (HM) of van der Linden (Psychometrika 72(3):287-308, 2007), the signed residual time model (SM) proposed by Maris and van der Maas (Psychometrika 77(4):615-633, 2012), and the drift diffusion model (DM) as proposed by Tuerlinckx and De Boeck (Psychometrika 70(4):629-650, 2005). One important distinction between these models is that the HM and the DM either assume or imply that accuracies and response times are independent given the latent trait variables, while the SM does not. In this paper we investigate the impact of this conditional independence property-or a lack thereof-on the manifest probability distribution for accuracies and response times. We will find that the manifest distributions of the latent trait models share several important features, such as the dependency between accuracy and response time, but we also find important differences, such as in what function of response time is being modeled. Our method for characterizing the manifest probability distributions is related to the Dutch identity (Holland in Psychometrika 55(6):5-18, 1990).

On Nonregularized Estimation of Psychological Networks

An important goal for psychological science is developing methods to characterize relationships between variables. Customary approaches use structural equation models to connect latent factors to a number of observed measurements, or test causal hypotheses between observed variables. More recently, regularized partial correlation networks have been proposed as an alternative approach for characterizing relationships among variables through off-diagonal elements in the precision matrix. While the graphical Lasso (glasso) has emerged as the default network estimation method, it was optimized in fields outside of psychology with very different needs, such as high dimensional data where the number of variables (p) exceeds the number of observations (n). In this article, we describe the glasso method in the context of the fields where it was developed, and then we demonstrate that the advantages of regularization diminish in settings where psychological networks are often fitted ( p≪n ). We first show that improved properties of the precision matrix, such as eigenvalue estimation, and predictive accuracy with cross-validation are not always appreciable. We then introduce nonregularized methods based on multiple regression and a nonparametric bootstrap strategy, after which we characterize performance with extensive simulations. Our results demonstrate that the nonregularized methods can be used to reduce the false-positive rate, compared to glasso, and they appear to provide consistent performance across sparsity levels, sample composition (p/n), and partial correlation size. We end by reviewing recent findings in the statistics literature that suggest alternative methods often have superior performance than glasso, as well as suggesting areas for future research in psychology. The nonregularized methods have been implemented in the R package GGMnonreg.

Exploring relapse through a network analysis of residual depression and anxiety symptoms after cognitive behavioural therapy: A proof-of-concept study

Objective: Many patients relapse within one year of completing effective cognitive behavioural therapy (CBT) for depression and anxiety. Residual symptoms at treatment completion have been demonstrated to predict relapse, and so this study used network analyses to improve specificity regarding which residual anxiety and depression symptoms predict relapse. Method: A cohort study identified relapse cases following low- and high-intensity CBT in a stepped care psychological therapy service. The sample included N = 867 "recovered" treatment completers that attended a six-month follow-up review. At follow-up, N = 93 patients had relapsed and N = 774 remained in-remission. Networks of final treatment session depression (PHQ-9) and anxiety (GAD-7) symptoms were estimated for both sub-groups. Results: Qualitatively similar symptom networks were found. Difficulty concentrating was a highly central symptom in the relapse network, whilst of only average centrality in the remission network. In contrast, trouble relaxing was highly central in the remission network, whilst of only average centrality in the relapse network. Discussion: Identification of central residual symptoms holds promise in improving the specificity of prognostic models and the design of evidence-based relapse prevention strategies. The small sample of relapse cases limits this study's ability to draw firm conclusions.

Exploring multicultural classroom dynamics: A network analysis

Students' relationships with peers and teachers strongly influence their motivation to engage in learning activities. Ethnic minority students, however, are often victimized in schools, and their educational achievement lags behind that of their majority group counterparts. The aim of the present study was to explore teachers' multicultural approach within their classrooms as a possible factor of influence over students' peer relationships and motivation. We utilized the novel methodology of estimating psychological networks in order to map out the interactions between these constructs within multicultural classrooms. Results indicate that a multicultural approach is directly connected to student motivation for both ethnic majority and minority students. Social integration within peer groups, however, seems to be a possible mediator of this relationship for the ethnic minority students. Due to the hypothesis generating nature of the psychological network approach, a more thorough investigation of this generated mediation hypothesis is called for.

Network structure of the Wisconsin Schizotypy Scales-Short Forms: Examining psychometric network filtering approaches

Schizotypy is a multidimensional construct that provides a useful framework for understanding the etiology, development, and risk for schizophrenia-spectrum disorders. Past research has applied traditional methods, such as factor analysis, to uncovering common dimensions of schizotypy. In the present study, we aimed to advance the construct of schizotypy, measured by the Wisconsin Schizotypy Scales-Short Forms (WSS-SF), beyond this general scope by applying two different psychometric network filtering approaches-the state-of-the-art approach (lasso), which has been employed in previous studies, and an alternative approach (information-filtering networks; IFNs). First, we applied both filtering approaches to two large, independent samples of WSS-SF data (ns = 5,831 and 2,171) and assessed each approach's representation of the WSS-SF's schizotypy construct. Both filtering approaches produced results similar to those from traditional methods, with the IFN approach producing results more consistent with previous theoretical interpretations of schizotypy. Then we evaluated how well both filtering approaches reproduced the global and local network characteristics of the two samples. We found that the IFN approach produced more consistent results for both global and local network characteristics. Finally, we sought to evaluate the predictability of the network centrality measures for each filtering approach, by determining the core, intermediate, and peripheral items on the WSS-SF and using them to predict interview reports of schizophrenia-spectrum symptoms. We found some similarities and differences in their effectiveness, with the IFN approach's network structure providing better overall predictive distinctions. We discuss the implications of our findings for schizotypy and for psychometric network analysis more generally.

Symptoms of social anxiety disorder and major depressive disorder: A network perspective

BACKGROUND

We used network analyses to examine symptoms that may play a role in the co-occurrence of social anxiety disorder (SAD) and major depressive disorder (MDD). Whereas latent variable models examine relations among latent constructs, network analyses have the advantage of characterizing direct relations among the symptoms themselves.

METHOD

We conducted network modeling on symptoms of social anxiety and depression in a clinical sample of 130 women who met criteria for SAD, MDD, both disorders, or had no lifetime history of mental illness.

RESULTS

In the resulting network, the core symptoms of social fear and depressed mood appeared at opposite ends of the network and were weakly related; so-called "bridges" between these symptoms appeared to occur via intervening variables. In particular, the worthless variable appeared to play a central role in the network.

LIMITATIONS

Because our data were cross-sectional, we are unable to draw conclusions about the direction of these effects or whether these variables are related to each other prospectively.

CONCLUSIONS

Continued testing of these pathways using longitudinal data will help facilitate the development of more effective clinical interventions for these disorders.

Network models of driver behavior

The way people behave in traffic is not always optimal from the road safety perspective: drivers exceed speed limits, misjudge speeds or distances, tailgate other road users or fail to perceive them. Such behaviors are commonly investigated using self-report-based latent variable models, and conceptualized as reflections of violation- and error-proneness. However, attributing dangerous behavior to stable properties of individuals may not be the optimal way of improving traffic safety, whereas investigating direct relationships between traffic behaviors offers a fruitful way forward. Network models of driver behavior and background factors influencing behavior were constructed using a large UK sample of novice drivers. The models show how individual violations, such as speeding, are related to and may contribute to individual errors such as tailgating and braking to avoid an accident. In addition, a network model of the background factors and driver behaviors was constructed. Finally, a model predicting crashes based on prior behavior was built and tested in separate datasets. This contribution helps to bridge a gap between experimental/theoretical studies and self-report-based studies in traffic research: the former have recognized the importance of focusing on relationships between individual driver behaviors, while network analysis offers a way to do so for self-report studies.

Beyond Clinical High-Risk State for Psychosis: The Network Structure of Multidimensional Psychosis Liability in Adolescents

OBJECTIVES

The main goal of the present study was to analyze the network structure of schizotypy dimensions in a representative sample of adolescents from the general population. Moreover, the network structure between schizotypy, mental health difficulties, subjective well-being, bipolar-like experiences, suicide ideation and behavior, psychotic-like experiences, positive and negative affect, prosocial behavior, and IQ was analyzed.

METHOD

The study was conducted in a sample of 1,506 students selected by stratified random cluster sampling. The Oviedo Schizotypy Assessment Questionnaire, the Personal Wellbeing Index-School Children, the Paykel Suicide Scale, the Mood Disorder Questionnaire, the Strengths and Difficulties Questionnaire, the Prodromal Questionnaire-Brief, the Positive and Negative Affect Schedule for Children Shortened Version, and the Matrix Reasoning Test were used.

RESULTS

The estimated schizotypy network was interconnected. The most central nodes in terms of standardized Expected Influence (EI) were 'unusual perceptual experiences' and 'paranoid ideation'. Predictability ranged from 8.7% ('physical anhedonia') to 52.7% ('unusual perceptual experiences'). The average predictability was 36.27%, implying that substantial variability remained unexplained. For the multidimensional psychosis liability network predictability values ranged from 9% (estimated IQ) to 74.90% ('psychotic-like experiences'). The average predictability was 43.46%. The results of the stability and accuracy analysis indicated that all networks were accurately estimated.

CONCLUSIONS

The present paper points to the value of conceptualizing psychosis liability as a dynamic complex system of interacting cognitive, emotional, behavioral, and affective characteristics. In addition, provide new insights into the nature of the relationships between schizotypy, as index of psychosis liability, and the role played by risk and protective factors.

The network of psychopathic personality traits: A network analysis of four self-report measures of psychopathy

Psychopathy is often perceived as a constellation of personality traits, yet there is little consensus as to what constitutes the core features of psychopathy. We applied a network approach to investigate the psychopathy network, as operationalized by four self-report measures of psychopathy among a large sample of undergraduate students. Items assessing manipulativeness and irresponsibility/impulsivity had the strongest centrality indices in the item-level psychopathy network models. Stimulus seeking, social deviance, and interpersonal/affective traits were the most central domains in the psychopathy network derived from all factors in the four psychopathy measures. Network analysis may offer an alternative approach to help researchers identify characteristics that are important in the psychopathy network. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Improved Insight into and Prediction of Network Dynamics by Combining VAR and Dimension Reduction

To understand within-person psychological processes, one may fit VAR(1) models (or continuous-time variants thereof) to multivariate time series and display the VAR(1) coefficients as a network. This approach has two major problems. First, the contemporaneous correlations between the variables will frequently be substantial, yielding multicollinearity issues. In addition, the shared effects of the variables are not included in the network. Consequently, VAR(1) networks can be hard to interpret. Second, crossvalidation results show that the highly parametrized VAR(1) model is prone to overfitting. In this article, we compare the pros and cons of two potential solutions to both problems. The first is to impose a lasso penalty on the VAR(1) coefficients, setting some of them to zero. The second, which has not yet been pursued in psychological network analysis, uses principal component VAR(1) (termed PC-VAR(1)). In this approach, the variables are first reduced to a few principal components, which are rotated toward simple structure; then VAR(1) analysis (or a continuous-time analog) is applied to the rotated components. Reanalyzing the data of a single participant of the COGITO study, we show that PC-VAR(1) has the better predictive performance and that networks based on PC-VAR(1) clearly represent both the lagged and the contemporaneous variable relations.

A Network Model of Resilience Factors for Adolescents with and without Exposure to Childhood Adversity

Resilience factors (RFs) help prevent mental health problems after childhood adversity (CA). RFs are known to be related, but it is currently unknown how their interrelations facilitate mental health. Here, we used network analysis to examine the interrelations between ten RFs in 14-year-old adolescents exposed ('CA'; n = 638) and not exposed to CA ('no-CA'; n = 501). We found that the degree to which RFs are assumed to enhance each other is higher in the no-CA compared to the CA group. Upon correction for general distress levels, the global RF connectivity also differed between the two groups. More specifically, in the no-CA network almost all RFs were positively interrelated and thus may enhance each other, whereas in the CA network some RFs were negatively interrelated and thus may hamper each other. Moreover, the CA group showed more direct connections between the RFs and current distress. Therefore, CA seems to influence how RFs relate to each other and to current distress, potentially leading to a dysfunctional RF system. Translational research could explore whether intervening on negative RF interrelations so that they turn positive and RFs can enhance each other, may alter 'RF-mental distress' relations, resulting in a lower risk for subsequent mental health problems.

A network analysis of female sexual function: comparing symptom networks in women with decreased, increased, and stable sexual desire

Problems related to low sexual desire in women are common clinical complaints, and the aetiology is poorly understood. We investigated predictors of change in levels of sexual desire using a novel network approach, which assumes that mental disorders arise from direct interactions between symptoms. Using population-based data from 1,449 Finnish women, we compared between-subject networks of women whose sexual desire decreased, increased, or remained stable over time. Networks were estimated and analyzed at T1 (2006) and replicated at T2 (2013) using R. Domains included were, among others, sexual functions, sexual distress, anxiety, depression, body dissatisfaction, and relationship status. Overall, networks were fairly similar across groups. Sexual arousal, satisfaction, and relationship status were the most central variables, implying that they might play prominent roles in female sexual function; sexual distress mediated between general distress and sexual function; and sexual desire and arousal showed different patterns of relationships, suggesting that they represent unique sexual function aspects. Potential group-differences suggested that sex-related pain and body dissatisfaction might play roles in precipitating decreases of sexual desire. The general network structure and similarities between groups replicated well; however, the potential group-differences did not replicate. Our study sets the stage for future clinical and longitudinal network modelling of female sexual function.

Symptoms of persistent complex bereavement disorder, depression, and PTSD in a conjugally bereaved sample: a network analysis

BACKGROUND

Complicated and persistent grief reactions afflict approximately 10% of bereaved individuals and are associated with severe disruptions of functioning. These maladaptive patterns were defined in Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) as persistent complex bereavement disorder (PCBD), but its criteria remain debated. The condition has been studied using network analysis, showing potential for an improved understanding of PCBD. However, previous studies were limited to self-report and primarily originated from a single archival dataset. To overcome these limitations, we collected structured clinical interview data from a community sample of newly conjugally bereaved individuals (N = 305).

METHODS

Gaussian graphical models (GGM) were estimated from PCBD symptoms diagnosed at 3, 14, and 25 months after the loss. A directed acyclic graph (DAG) was generated from initial PCBD symptoms, and comorbidity networks with DSM-5 symptoms of major depressive disorder (MDD) and post-traumatic stress disorder (PTSD) were analyzed 1 year post-loss.

RESULTS

In the GGM, symptoms from the social/identity PCBD symptoms cluster (i.e. role confusion, meaninglessness, and loneliness) tended to be central in the network at all assessments. In the DAG, yearning activated a cascade of PCBD symptoms, suggesting how symptoms lead into psychopathological configurations. In the comorbidity networks, PCBD and depressive symptoms formed separate communities, while PTSD symptoms divided in heterogeneous clusters.

CONCLUSIONS

The network approach offered insights regarding the core symptoms of PCBD and the role of persistent yearnings. Findings are discussed regarding both clinical and theoretical implications that will serve as a step toward a more integrated understanding of PCBD.

Network analysis: a brief overview and tutorial

Objective : The present paper presents a brief overview on network analysis as a statistical approach for health psychology researchers. Networks comprise graphical representations of the relationships (edges) between variables (nodes). Network analysis provides the capacity to estimate complex patterns of relationships and the network structure can be analysed to reveal core features of the network. This paper provides an overview of networks, how they can be visualised and analysed, and presents a simple example of how to conduct network analysis in R using data on the Theory Planned Behaviour (TPB). Method: Participants (n = 200) completed a TPB survey on regular exercise. The survey comprised items on attitudes, normative beliefs, perceived behavioural control, and intentions. Data were analysed to examine the network structure of the variables. The EBICglasso was applied to the partial correlation matrix. Results: The network structure reveals the variation in relationships between the items. The network split into three distinct communities of items. The affective attitude item was the central node in the network. However, replication of the network in larger samples to produce more stable and robust estimates of network indices is required. Conclusions: The reported network reveals that the affective attitudinal variable was the most important node in the network and therefore interventions could prioritise targeting changing the emotional responses to exercise. Network analysis offers the potential for insight into structural relations among core psychological processes to inform the health psychology science and practice.

The Gaussian Graphical Model in Cross-Sectional and Time-Series Data

We discuss the Gaussian graphical model (GGM; an undirected network of partial correlation coefficients) and detail its utility as an exploratory data analysis tool. The GGM shows which variables predict one-another, allows for sparse modeling of covariance structures, and may highlight potential causal relationships between observed variables. We describe the utility in three kinds of psychological data sets: data sets in which consecutive cases are assumed independent (e.g., cross-sectional data), temporally ordered data sets (e.g., n = 1 time series), and a mixture of the 2 (e.g., n > 1 time series). In time-series analysis, the GGM can be used to model the residual structure of a vector-autoregression analysis (VAR), also termed graphical VAR. Two network models can then be obtained: a temporal network and a contemporaneous network. When analyzing data from multiple subjects, a GGM can also be formed on the covariance structure of stationary means-the between-subjects network. We discuss the interpretation of these models and propose estimation methods to obtain these networks, which we implement in the R packages graphicalVAR and mlVAR. The methods are showcased in two empirical examples, and simulation studies on these methods are included in the supplementary materials.

Personalized Network Modeling in Psychopathology: The Importance of Contemporaneous and Temporal Connections

Recent literature has introduced (a) the network perspective to psychology and (b) collection of time series data to capture symptom fluctuations and other time varying factors in daily life. Combining these trends allows for the estimation of intraindividual network structures. We argue that these networks can be directly applied in clinical research and practice as hypothesis generating structures. Two networks can be computed: a temporal network, in which one investigates if symptoms (or other relevant variables) predict one another over time, and a contemporaneous network, in which one investigates if symptoms predict one another in the same window of measurement. The contemporaneous network is a partial correlation network, which is emerging in the analysis of cross-sectional data but is not yet utilized in the analysis of time series data. We explain the importance of partial correlation networks and exemplify the network structures on time series data of a psychiatric patient.

Replicability and Generalizability of Posttraumatic Stress Disorder (PTSD) Networks: A Cross-Cultural Multisite Study of PTSD Symptoms in Four Trauma Patient Samples

The growing literature conceptualizing mental disorders like posttraumatic stress disorder (PTSD) as networks of interacting symptoms faces three key challenges. Prior studies predominantly used (a) small samples with low power for precise estimation, (b) nonclinical samples, and (c) single samples. This renders network structures in clinical data, and the extent to which networks replicate across data sets, unknown. To overcome these limitations, the present cross-cultural multisite study estimated regularized partial correlation networks of 16 PTSD symptoms across four data sets of traumatized patients receiving treatment for PTSD (total N = 2,782). Despite differences in culture, trauma type, and severity of the samples, considerable similarities emerged, with moderate to high correlations between symptom profiles (0.43-0.82), network structures (0.62-0.74), and centrality estimates (0.63-0.75). We discuss the importance of future replicability efforts to improve clinical psychological science and provide code, model output, and correlation matrices to make the results of this article fully reproducible.

Evidence that psychopathology symptom networks have limited replicability

Network analysis is quickly gaining popularity in psychopathology research as a method that aims to reveal causal relationships among individual symptoms. To date, 4 main types of psychopathology networks have been proposed: (a) association networks, (b) regularized concentration networks, (c) relative importance networks, and (d) directed acyclic graphs. The authors examined the replicability of these analyses based on symptoms of major depression and generalized anxiety between and within 2 highly similar epidemiological samples (i.e., the National Comorbidity Survey-Replication [n = 9282] and the National Survey of Mental Health and Wellbeing [n = 8841]). Although association networks were stable, the 3 other types of network analysis (i.e., the conditional independence networks) had poor replicability between and within methods and samples. The detailed aspects of the models-such as the estimation of specific edges and the centrality of individual nodes-were particularly unstable. For example, 44% of the symptoms were estimated as the "most influential" on at least 1 centrality index across the 6 conditional independence networks in the full samples, and only 13-21% of the edges were consistently estimated across these networks. One of the likely reasons for the instability of the networks is the predominance of measurement error in the assessment of individual symptoms. The authors discuss the implications of these findings for the growing field of psychopathology network research, and conclude that novel results originating from psychopathology networks should be held to higher standards of evidence before they are ready for dissemination or implementation in the field. (PsycINFO Database Record