While the weighting procedure above allowed us to highlight the importance of significant information results, in time bins where few significant information results were observed, an upwards bias in the information results would still be observed. To detect cases where the ensemble of information values was not significantly different from null, we also used a Kolmogorov–Smirnov test to compare the distribution of real MI results to the distribution of MI results from null surrogate data used to calculate the individual neuron p values. This allowed us to assess the time bins for which the entire ensemble of neurons was not significantly different from null data, suggesting the ensemble as a whole was not encoding significant amounts of information (open circles in figures). We applied a threshold of p < 0.01 to all such Kolmogorov–Smirnov tests to assess significant ensemble encoding.
