Figure 2.

Data analysis methods. From a dataset we derived seven statistical measures of phase opposition. A time-frequency transform was used to extract oscillatory phase for each trial, time point, and frequency. Inter-trial coherence values were then computed for each trial group (outcome A vs. B) as well as for the entire dataset (both outcomes pooled). Using these three ITC-values, the circular Watson-Williams parametric test directly yielded a time-frequency map of p-values (top-right, blue time-frequency map). The same three ITC-values were used to calculate time-frequency maps of the phase opposition measures POS, POP, and PBI. To determine the statistical significance of these measures, two distinct non-parametric procedures were applied. The first consisted in randomly permuting the trial labels (assignment of outcome A vs. B), and for each permutation recalculating POS, POP, and PBI. The ranking (percentile) of the original dataset values against these null-hypothesis distributions could be used as a p-value (right hand-side, green time-frequency maps). Alternately, the null-hypothesis distributions could be summarized by their mean and standard deviation (across permutations), the values of the original dataset could then be expressed as a standardized z-score against the null distribution, and a p-value assigned using the normal cumulative distribution function (right hand-side, red time-frequency maps).