This is the first study to delve deeply into the symptomatic relationships within depression and anxiety disorders and to estimate the complex relationship of the depression-worry & meta-worry and anxiety-worry & meta-worry mechanisms with psychological networks. While "Lack of cognitive self-confidence Negative beliefs about uncontrollability and danger" sub-scales of the meta-cognition questionnaire, and "My worries overwhelm me" item of the PSWQ (worry) showed comorbidity with both depression and anxiety symptoms. Proximity takes the inverse of the sum of all the shortest paths between one node and all other nodes in the network.
Therefore, detection of the most important symptom is crucial and can be calculated via centrality measures. The first goal of this study is to explore depression and anxiety symptoms to discover core symptoms and highly correlated symptoms of each. Third aim is to find out how worry and meta-worry concepts interact with depression and anxiety symptoms.
The fourth aim is to observe where elements of worry and meta-worry fit in the causal relationship between symptoms of depression and anxiety. After uncovering the significance of each feature of the depression and anxiety measures, the correlations of these individual entities with the worry and meta-worry items are analyzed.
Datasets
Measures
Instead of each item of the questionnaire, only the subscale scores were included in the network analysis. To avoid items measuring the same in the network, only 8 items were included in the network analysis. To ensure theoretical support, item selection was done based on the short version of the PSWQ.
Participants
Networks
In the first network, only depression symptoms (BDI-II items) were included, therefore 5015 participant data were used to estimate this. Similarly, in the second network, only the anxiety symptoms (BAI items) were used with 5430 participant data. The bridge centrality is calculated for each node connecting to each node in the other community.
Here again, A and C are conditionally independent of B. Finally, in the collision case, A and C together result in a third variable, B. In the first directed network, only depressive symptoms (item BDI-II) were included, so 5015 participants' data were used to estimate . Similarly, another directional network used only anxiety symptoms (BAI items) with data on 5430 participants.
If in at least 51% of the networks an edge from symptom X pointed to symptom Y, then this direction was reflected in the final, average network. Second, a DAG where the edge thickness indicates the probability that the edge is pointing in the direction.
Regularized partial correlation networks
However, in the centrality stability diagram shown in Figure 7, it can be observed that proximity stability is lower than strength and expected influence, which are close to the upper limit. The bottom graph of the same figure shows the excellent edge weight stability plotted through the non-parametric bootstrap. These nodes were found to be significantly different from the remaining nodes (see Figure 9 for more details).
The expected influence measure supports the same result as the strength measure, as there were no apparent red edges in the network. Although the stability of the betweenness and closeness measures were acceptable and good, respectively, they did not provide any centrally strong nodes. In the Figure 10A, there is no recognizable clustering between nodes, which means that latent variable models may not fully explain the complexity of the relationship between depression symptoms in given survey.
In the previous studies that demonstrated the validity and reliability of the Korean version of the BDI-II survey, factor analysis was also reported at the same time. Figure 10C shows that the two-factor structure best fits the results of Song et al. This adjustment appears to be a complete miss in terms of representing the current data. [27] Finally, Figure 10D shows the three-factor structure reported by Park et al.
Centrality stability coefficients calculated by the bootstrap were lowest for betweenness at 0.57, which is still considered good. Similar to the stability of the depression network, the centrality stability plot in Figure 12 indicates that the stability of closeness is lower than the strength and expected impact, which is close to the ceiling limit. The lower part of the same figure shows excellent edge weight stability that was plotted through the non-parametric bootstrap.
It was revealed that these nodes are significantly different from the remaining nodes (See Figure 14 for more details). In Figure 15A, in contrast to the depression network, there appears to be three discernible clustering between the nodes. In Figure 14B, different colors belong to four different factor structures which according to Lee et al.
Bridge Analysis
In the first network, depression, worry and meta-cognition nodes/items are included (Figure 16). Similar to the single measure GLASSO networks above, the stability of the network must be good to be able to go forward and interpret the results. Centrality stability coefficients calculated by the case dropstart for betweenness which was 0.20 and closeness 0.36 were within the acceptable range to establish the network as stable.
The excellent stability of the edge weight, which was plotted via the non-parametric bootstrap, can also be seen in the same figure. In interpreting the results, unlike the single GlASSO networks, it was not the most central nodes or strongest edges that were important, but rather the bridge nodes connecting two communities. Due to the lower stability of the bridge intermediateness and the bridge proximity together with the lack of the number of red edges and the dimensions, which leads to the expected influence of the bridge being almost identical to the strength of the bridge, only the measure for bridge strength was examined (Figure 18). ).
The correlation matrix between all nodes was checked to determine which node in the depressed community they were most associated with (Figure 20). Anxiety, worry and metacognition nodes/items were included in the second estimated network (Figure 21). Centrality stability coefficients calculated by bootstrapping leaving out the case for a betweenness of 0.13 were very low, making this measure of centrality uninterpretable.
Strength and expected influence scored 0.67 and closeness 0.52 which is considered in the good range for stability analysis. Others fell into the stable good category; bridge strength and bridge expected influence with scores of 0.67 and bridge betweenness with a score of 0.52.(Figure 22). It also shows excellent edge weight stability plotted by the non-parametric bootstrap in the same figure.
34;(M1)Lack of cognitive confidence", (M4)Negative beliefs about uncontrollability and danger" and "(W2)My worries overwhelm me" nodes showed the highest bridge strength from community of concern and metacare, while. The correlation matrix between all nodes was checked to discover the strongest associations of nodes M1, M4 and W2 with the nodes of the anxiety community (Figure 25). However, no bridging nodes from the depressed side of the network were highly central nodes.
Directed Acyclic Graph (DAG)
Meanwhile, "(W2)My worries overwhelm me" only contributed to the lower branches of the depression symptoms. In figure 34, "(M4)Negative beliefs about uncontrollability and danger" and "(W2)My worries overwhelm me" bridge symptoms for anxiety appear to be contributors to the main anxiety symptoms. Meanwhile, "(M5)Need to control thoughts" only contributed to the lower branches of the anxiety symptoms.
The main focus of this dissertation was to investigate four objectives explained in the objective of the study above. The first two objectives of the study were successfully achieved through a detailed analysis of the relationships between depression and anxiety symptoms. When the strongest association with the symptoms was checked, these did not appear to be one of the central symptoms.
When causality was examined with optimal cut-off DAG, it had an impact on the initiating symptoms of anxiety, while only contributing to the lower branches of the depression symptoms. Their strongest association with anxiety symptoms included one of the central anxiety symptoms, namely "Shaky/Unsteady". The period in which the data was collected, historical, economic or sociological events that could potentially impact the mental health of the population in question, and the effect of a recent pandemic.
Cramer, “State of the art personality research: a tutorial on network analysis of personality data in R,” Journal of Research in Personality, vol. Choi, “Diagnostic utility and psychometric properties of the Beck Depression Inventory-II among Korean adults,” Frontiers in Psychology, vol. Kim, “Validation and Factor Structure of the Korean Version of the Beck Depression Inventory Second Edition (BDI-II): In a Sample of University Students,” Korean Journal of Biological Psychiatry, pp.
Lee, “Reliability and validity of the Korean version of the Beck Depression Inventory-ii via the Internet: Results from a university student sample,” Korean Journal of Biological Psychiatry , vol. 30] Han-Kyeong Lee, Jihae Kim, SanghwangHong, Eun-ho Lee, and Hwang Soon Taeg, "Psychometric properties of the Beck Anxiety Inventory in a community-dwelling sample of Korean adults," Korean Journal of Clinical Psychology, vol. Borkovec, "Development and Validation of the Pen State Concern Questionnaire," Behavior Research and Therapy, vol.
Kwon, “The Penn State Worry Questionnaire: Psychometric Properties of the Korean Version,” Depression and Anxiety, vol. Bühlmann, “High-dimensional graphs and variable selection with the lasso,” The annals of Statistics, vol.
