Fig. 3

Single-trial SVM classification performance for ≤ 42 vs. > 42 NVIQ FXS subgroups. Each bar represents the percent correct classification achieved with each feature. The features represented on the x-axis are single-trial repetition suppression-induced EEG energy modulations between two presentations of the same stimulus, computed within a 0- to 800-ms window (total number of observations for each feature n = 252, but 216 were used to ensure balanced classes using bootstrapping; see the ‘Methods’ section for details). The highest decoding (65.2%) was found with FR 1–4, i.e. energy at the right frontal region between presentations P1 and P4. The y-axis starts at the theoretical chance level of 50%. The horizontal lines represent respectively (from bottom to top) the chance levels using binomial cumulative distribution for p < 0.05 and p < 0.01, corrected for multiple comparisons across all 119 features. The error bars represent the standard error on the mean (s.e.m) computed across the bootstrap repetitions. C central, FC fronto-central, FR frontal-right, TL temporal-left