Plotting the time-course of β coefficients for a given predictor indicates when in the trial the predictor explains variance. Separate regression analyses were run for each subject; traces show cross-subject mean, shaded areas cross-subject standard error. Dots above the traces show time-points where the β coefficient was significantly different from 0, assessed using a t test on the cross-subject distribution with Benjamini-Hochberg correction for comparison of multiple timepoints. In (D), the regression analysis included only a single predictor coding for the trial outcome. This regression has significant positive coefficients for the outcome predictor before information about the outcome is available (red shaded area). This is because the subjects’ expectation of reward drives variation in the signal which is correlated with the trial outcome (as seen in B).
