Skip to main content
Fig. 6. | Biology of Sex Differences

Fig. 6.

From: Univariate and multivariate sex differences and similarities in gray matter volume within essential language-processing areas

Fig. 6.

Structure of the LR models fitted in the raw and PCP datasets. A and B Nomograms illustrating the relative contribution of each component to the SENT_CORE network to the Pclass scores yielded by the LR models fitted in the raw (reference sex category: males) and in the PCP (reference sex category: females) datasets. The values of three discrimination indexes (R2, C index, and Somers’ D) of each of these two models are reported within the plots. Although nomograms are ordinarily used to predict individual classification probabilities, in this case, the males (blue) and females (red) medians are used to represent how the scores of these groups in each feature were scored (i.e., points; orange numbers), additively integrated in composites (“total points”), and non-linearly project to the Pclass continuum on which the multivariate sex differences and similarities displayed in Fig. 4 were estimated. Note that to enhance readability: (1) brain features are decreasingly sorted according to their contribution to the model; (2) instead of including a points’ axis, the points achievable (orange numbers) in each scale are represented back-to-back to the features’ values (black numbers); (3) the marks of some scales have been suppressed; and, (4) to highlight them, the scales of those features achieving statistical significance are depicted with thicker lines. C Ordinal relationship (quantified through the Spearman’s rho correlation index) between the regression coefficient values and the size of the univariate differences (medians’ difference) in the raw (left) and PCP (right) datasets. Note that the sign of this association is largely arbitrary as it arises from the different sex category used as reference in the raw and PCP models. D Ordinal relationship (quantified through the Spearman’s rho correlation index) between the coefficient values of the LR models fitted in the raw and PCP datasets. To ease the visualization of the relationships depicted in panels C and D, trend lines obtained through gam-smoothing (and their 95% interval; yellow shade) have been added

Back to article page