Skip to main content
. Author manuscript; available in PMC: 2024 Feb 19.
Published in final edited form as: Neuroimage. 2022 Jul 14;260:119457. doi: 10.1016/j.neuroimage.2022.119457

Fig. 10.

Fig. 10.

Performance (accuracy) of the classifier at 200 msec for different feature sets. For each model, the maximum performance across a range of C and 𝜆 (free parameters of the SVM) was saved and averaged across subjects. Error bars represent SEM. Purple bars on the left of each panel show performance of classifiers trained on the ERP only for a range of window lengths. In all three window lengths, the features consisted of the mean, linear slope, quadratic and cubic over that time window. In the right portion of each panel is shown a 2 × 5 analysis crossing the mean power in five frequency bands and whether the ERP was also included. The time-window was always 250 msec and the ERP (if present) was coded as the mean, linear slope and quadratic within the time -window.