Figure 4.

Model-based validation of α estimates. A, We used the estimated stimulus-processing model for each neuron to validate our nonparametric estimates of α. Specifically, we computed α directly from the model-predicted firing rates on each trial. These model-based estimates of α were in close agreement with nonparametric estimates across the population of neurons (Spearman's ρ = 0.92; p = 7.5 × 10⁻⁴⁰). B, This close agreement was despite the fact that the models typically captured about half of the stimulus-driven firing rate variance (median: 0.54, interquartile range: 0.37–0.68). Arrow indicates the median.