Figure 1.
Scree-plots of facet level (A) and item level (B) principal component analyses.
A. At facet level, a 6-factor structure is suggested by the screeplot. B. At item level, a 10-factor structure is found without any resemblance to either a 5-factor structure or a 30-facet structure. Such item level decompositions are known to be unreliable in smaller-than-standard datasets such as the present dataset. Hence, both 5-factor (‘confirmatory’) and 6-factor (‘exploratory’) PCAs were performed, to force item level results to more plausible solutions. The factor structures of these PCAs, rather than the 10-factor structure, served as templates in the NCS-to-FS matching procedure. See text for further details.
Table 1.
The results of standard 5-factor, confirmatory 5-factor and exploratory 6-factor PCAs of our dataset.
Figure 2.
Results of the network community structure to factor structure matching procedure at facet level.
A. Results of the NCS-to-FS matching procedure for standard, confirmatory and exploratory FSs. X-axis shows the correlation coefficient r as a threshold for significance of a link in the network graph (as r increases to the right, more links are pruned from the network). Y-axis shows normalized dissimilarity (mismatch) scores. Blue: standard 5-FS, red: confirmatory 5-FS, green: exploratory 6-FS. The confirmatory 5-FS shows the best match with NCS at r = 0.271, p = 4.89E-09. For details, see text and Tables 1 and 2. B. Results of the NCS-to-FS matching procedure for the specific case of the winning confirmatory 5-FS (red line in Fig A), with a subspecification of the matching results per factor. F1–F5: confirmatory factors resembling Neuroticism, Extraversion, Openness, Agreeableness and Conscientiousness, respectively. For details, see text and Tables 1 and 2.
Table 2.
Table showing the quantitative results of the cluster-to-factor matching procedure at facet level.
Figure 3.
The Personality Web at facet level: network graph of correlational relationships between the 30 facets of the NEO-PI-R.
The community structure of this graph has an overall best fit with the confirmatory 5-FS, occurring at r>0.271, p<4.89 E-09. See Table 2 for significances and correlation coefficients. Node = facet, link = significant correlation. Red links: positive correlations. Blue links: negative correlations. The thickness of the links represents the strength of the correlation. For further information, see Table 4B. n = neuroticism, e = extraversion, o = openness, a = agreeableness, c = conscientiousness. Numbers refer to facet number. Nodes are positioned in clusters according to their factor membership (standard 5-FS). The color of nodes denotes their network cluster membership. Only two facets show a mismatch with the standard 5-FS (n5 and n2). Both mismatches involve the neuroticism dimension depicted below in red. These facets have strong correlations with facets from the conscientiousness cluster (blue) and the agreeableness cluster (green), as can be observed by the thickness of the corresponding links. As a result, n5 is “drawn” into the conscientiousness cluster and n2 into the agreeableness cluster.
Figure 4.
Results of the network community structure to factor structure matching procedure at item level.
A. Results of the NCS-to-FS matching procedure for standard, confirmatory and exploratory factor analyses. Same specifications as in Figure 2 apply. Minimal dissimilarity (mismatch) is found with the standard 5-FS at r = 0.164, p = 3.08E-04. For details, see text and Tables 1 and 3. B. Results of the NCS-to-FS matching procedure for the specific case of the winning standard 5-FS (blue line in Figure A), with a subspecification of the matching results per factor. N, E, O, A,C: NEUROTICISM, EXTRAVERSION, OPENNESS, AGREEABLENESS and CONSCIENTIOUSNESS. Different factors show various degrees of matching with NCSs. For details, see text and Tables 1 and 3.
Table 3.
Table showing the quantitative results of the cluster-to-factor matching procedure at item level.
Figure 5.
The Personality Web at item level: network graph of correlational relationships between the 240 items of the NEO-PI-R.
The community structure of this graph has an overall best fit with the standard 5-FS, occurring at r = 0.164, p = 3.08E-04. Node = item, link = significant correlation. Red links: positive correlations. Blue links: negative correlations. Node size = degree (larger nodes are bigger hubs, scale = 1 to 10). For further information, see supporting information (Table S2). The color of the items refers to their standard factor membership, i.e. red: Neuroticism, orange: Extraversion, yellow: Openness, green: Agreeableness, blue: Conscientiousness. Items are grouped together according to their network cluster membership, and clusters are depicted as circles. Clusters are grouped according to size. Five clusters are found that show maximum correspondence to standard factors (Table 3,4). A small 6th factor is found (right in the graph), consisting of 9 items belonging to extraversion (1 item), openness (3 items), agreeableness (3 item) and neuroticism (2 items). Four isolates (bottom-right corner) consisting of 1 or 2 items were discarded from further analysis. Most of the top 10% hubs are located in the extraversion and neuroticism clusters.
Figure 6.
The Personality Web at cluster level: network graph showing correlations between cluster scores calculated using the network community structure from Figure 5.
CLUSTER1_N, CLUSTER2_E, CLUSTER3_O, CLUSTER4_A, CLUSTER5_C: clusters showing maximum correspondence with standard clusters of Neuroticism, Extraversion, Openness, Agreeableness and Conscientiousness, respectively. CLUSTER6: newly found sixth factor (see Figure 5, Table 4 and Discussion). Network graph is shown at p<0.01 corrected for multiple comparisons. The thickness of the links represents the strength of the correlation coefficient. Correlation coefficients are shown alongside the links. Red: positive correlations, blue: negative correlations. For further information, see Table 4A. CLUSTER1_N, the Neuroticism analogue, has only negative (inhibitory) influences on the remaining network structure, which shows only positive interrelations. Personality Webs such as these may represent developmental structures. CLUSTER2_E (Extraversion analog), CLUSTER5_C (Conscientiousness analog) and the 6th cluster form an intermediate structure in between CLUSTER1_N (Neuroticism analog) and CLUSTER4_A and CLUSTER3_O (the Agreeableness and Openness analogs). This suggests that the negative influence of Neuroticism on Agreeableness and Openness scores, as observed in personality disorders, is mediated by this intermediate level (see discussion for further details).
Table 4.
Cluster contents and node-specific network metrics for the network graphs described in Figures 3 and 6.
Table 5.
Overall network metrics for the ‘winning’ network graphs described in this paper.