Fig. 6: Comparison of areas under the receiver operating characteristics curve (AUCs) for entropy (red) and spectral (blue) features selected using linear mixed models (LMMs). | Communications Biology

Fig. 6: Comparison of areas under the receiver operating characteristics curve (AUCs) for entropy (red) and spectral (blue) features selected using linear mixed models (LMMs).

From: Neural complexity is a common denominator of human consciousness across diverse regimes of cortical dynamics

Fig. 6: Comparison of areas under the receiver operating characteristics curve (AUCs) for entropy (red) and spectral (blue) features selected using linear mixed models (LMMs).The alternative text for this image may have been generated using AI.

As in Fig. 5, we performed three basic comparisons (a fcEntropy vs fcSpectral; b scEntropy vs scSpectralA; c scEntropy vs scSpectralR) with EEG data from three datasets [AS (top row) training data; NT (middle row) validation data; Dup15q (bottom row) validation data]. In each case, we generated 10,000 bootstrapped resamples to derive AUC confidence intervals in Supplementary Table 3. Note that the number of resamples on the vertical axis of each histogram is log-scaled due to the very large range of resamples across histogram bins. Also note that classifier performance was saturated using entropy features for all comparisons except for aiii (Dup15q FC features), i.e., all bootstrapped resamples yielded 100% AUC for entropy features in these instances. Five out of the nine comparisons above yield significantly larger AUCs for entropy features, as indicated by asterisks. Source data are presented in Supplementary Data 5.

Back to article page