Extended Data Fig. 2: NPvC goodness of fit for different cell types.

a, Schematic of the c-vine copula sequential computation of four-dimensional data. The multivariate copula density function is computed from the product of a series of sequential bivariate copulas, which are being computed as it is explained in Supplementary Note ‘Vine copula modeling of neural responses’. b, Distribution of virtual heading direction, lateral velocity, lateral acceleration, forward velocity, and forward acceleration during the trial for different trial types. c, The cumulative distribution of goodness of fit (fraction of deviance explained) for NPvC model for different populations. d, Mutual information between task variables (sample cue, test cue, choice, and reward direction) and behavioral variables (virtual heading direction, lateral velocity and acceleration, and forward velocity and acceleration). In b and d, lines and shaded areas plot mean and s.e.m. over experimental sessions.