Table 3 Relationships between structural and semantic graph features and psychosis.

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

Graph feature

Task

Static action-predication graph (S_AP)

Dynamic action-predication graph (D_AP)

Static sequential graph (S_Seq)

Dynamic sequential graph (D_Seq)

A. Size

Number of nodes (NN)

Picture

0.17*

0.27***

0.20*

0.34***

Open-ended

0.13

0.21*

0.14

0.33***

Number of edges (NE)

Picture

0.19*

0.31***

0.17*

0.27**

Open-ended

0.12

0.19*

0.11

0.37***

Diameter

Picture

0.16

0.23**

0.24**

0.30***

Open-ended

0.00

0.03

0.28***

0.25**

Average shortest path length (ASPL)

Picture

0.11

0.18*

0.24**

0.28**

Open-ended

0.08

0.17*

0.27**

0.26**

B. Connectedness

Average weighted degree (AWD)

Picture

0.15

0.22**

−0.05

−0.32***

Open-ended

−0.03

0.06

−0.13

−0.29***

Density

Picture

−0.06

−0.15

−0.26**

−0.33***

Open-ended

−0.20*

−0.23**

−0.21*

−0.31***

Size of largest strongly connected component (LSCC)

Picture

0.15

0.24**

0.20*

0.18*

Open-ended

0.08

0.13

0.14

0.29***

C. Organization

LSCC z-score (LSCCZ)

Picture

-0.22**

−0.31***

0.28***

0.25**

Open-ended

−0.09

−0.13

0.26**

0.31***

ASPL z-score (ASPLZ)

Picture

−0.22**

−0.30***

−0.02

0.25**

Open-ended

−0.11

−0.14

0.01

0.24**

  1. Graph features are categorized into three domains of size, connectedness, and psychosis. For each feature rank biserial correlation coefficient (RBC) is reported for picture description task (Picture) and open-ended narrative (Open-ended) as a measure of effect size. The significant associations are tagged by asterisks (*<0.05, **<0.01, and ***<0.001). Associations survived in Bonferroni correction are bolded (alpha = 0.0003). Semantic graphs showed a preference for picture description tasks with higher effect sizes and more significant associations. Structural graphs showed a task-independent relation with psychosis. In overall, dynamic graph features outperformed static counterparts in both graph types.
Back to article page