Fig. 2: Correlations between structural and semantic graph features and dimensional clinical characteristics. | Schizophrenia

Fig. 2: Correlations between structural and semantic graph features and dimensional clinical characteristics.

From: Who does what to whom? graph representations of action-predication in speech relate to psychopathological dimensions of psychosis

Fig. 2

a Heatmap representations of the Spearman’s correlation coefficient for structural and semantic graph features and clinical measures in picture description and open-ended narrative tasks across all participants. Significant relationships with uncorrected p values < 0.05 are shaded based on their effect sizes (Spearman’s rho). Correlations surviving Bonferroni correction are starred. Bar plots of correlation coefficients per clinical dimension are available in Supplementary Fig. 2. b Network representation of significant relationships between graph features and clinical measures. Multi-collinearities were separately handled for structural and semantic graph features by stepwise comparison of variance inflation factors and feature exclusion. Multiple comparisons were accounted for using Bonferroni correction. S_AP static action-predication graph feature, D_AP dynamic action-predication graph feature, S_SEQ static sequential graph feature, D_SEQ dynamic sequential graph feature, NN number of nodes, NE number of edges, diameter graph diameter, ASPL average shortest path length, AWD average weighted degree, density graph density, LSCC size of largest strongly connected component, LSCCZ z-score of LSCC compared to 1000 random graphs, ASPLZ z-score of ASPL compared to 1000 random graphs. More details on graph features are available in Supplementary Table 3.

Back to article page