Fig. 3: Different aspects of pre-stimulus features contributed to the modulation model.
a From left to right, bar plot for the averaged increasement classification d’ from post-stimulus only classification across all electrodes for: (all) including all pre-stimulus features, (stBHA) including the pre-stimulus stBHA features only, (stP) including the pre-stimulus stP features only, (Ph) including the pre-stimulus phase features only, (stBHA+stP) including the pre-stimulus stBHA and stP features, (stBHA+Ph) including pre-stimulus stBHA and phase features, (stP+Ph) including the pre-stimulus stP and phase features. Each sample dot in the plot represents the averaged value from one single electrode, n = 246 samples in total for each plot; data are presented as mean values ± s.e.m.; *** p < 0.001, two-sided paired t-test, Bonferroni correction for multiple repeated tests. (From left to right, p = 6 × 10−17, 3 × 10−29, 4 × 10−25, 7 × 10−18, 2 × 10−20, 4 × 10−19, 8 × 10−19, respectively). b From left to right, bar plot for distributions of: (1) the averaged z-scored stBHA power, (2) the standard deviation of z-scored stBHA power, (3) the averaged absolute value of stP, (4) the standard deviation of stP, within [−500−ms,−100 ms] pre-stimulus time window for low MI and high MI trials in each electrode. Each sample dot in the plot represents the averaged value from one single electrode, n = 246 samples in total for each plot; data are presented as mean values ± s.e.m., *** p < 0.001, permutation test. (From left to right, p = 7 × 10−6, 2 × 10−8, 3 × 10−23, 4 × 10−16, respectively. c The averaged empirical probability of having non-zero weights in the sparse GLM model for different pre-stimulus phase features of different frequency. Shaded area indicates 95% confidence interval under random feature selection. (Source data are provided as a Source Data file.).
