Regression analyses of population activity (Figures 3E–3H and 5) comprised a set of linear regressions each of which predicted the log2 transformed activity of one neuron at one time point relative to trial events. For each neuron-time point we calculated the coefficient of partial determination (CPD) for each predictor, i.e., how much variance of the neurons activity at that time-point was explained by the full regression analysis that was not explained by the regression analysis if that predictor was removed. This is a measure of how much variance is uniquely explained by a predictor that cannot be explained by the other predictors. To assess how much variance of the population activity was explained by a given predictor at a given time point, we averaged the CPDs for all neurons at that time point to yield the population CPD time-courses shown in Figures 3E and 5.
