Extended Data Fig. 3: Replicating the main findings for photo-identified and non-photo-identified putative DANs.

a–e, Analysis from Fig. 3 with n = 122 photo-identified and non-photo-identified putative DANs. b, Mean correlation coefficient r=0.91 and geometric mean of the two-tailed t-test p-values (p-value=6.57e-48, n=10,000 partitions). Inset: the histogram of the regression slopes, 95% CI=(0.99,1.01). The error bars in c represent the 99% CI of the cross validation of temporal discount factors of single neurons using 50% of the total number of trials per delay n = 10,000 times. f, Estimated single-neuron temporal discount factors as a function of the position in the ML axis. Dots are colour-coded by mouse identity and dots with a blue outer circle correspond to the photo-identified neurons. The line represents the Huber Regression fit. Spearman rank correlation over neurons, two-tailed test p-value=0.26, n=86. g,h,k, Analysis from Fig. 4 with n = 122 photo-identified and non-photo-identified putative DANs. g, Pearson correlation coefficient over neurons r=0.2, two-tailed test n=98, p-value=0.0486, 95% CI=(-0.007,0.38). i,j,l,m, Analysis from Fig. 5 with n = 122 photo-identified and non-photo-identified putative DANs. i, Pearson correlation coefficient over n=98 neurons r=-0.65, two-tailed t-test p-value=1.14e-13, 95% CI=(-0.75,-0.52). j, Bootstrapped 10,000 times to test if variance of corrected firing rate for variable cue is greater than for certain cue: one-tailed p-value=0.093, n=122 neurons, mean difference in the variances=0.020, 95% CI=(-0.0053,0.089). l, Bootstrapped 10,000 times testing if update in temporal discount when removing the shortest or longest delay is different over neurons: two-tailed p-value=0.07, n=122, mean absolute difference=0.04, 95% CI=(0.0015, 0.11). m, Pearson correlation coefficient over neurons r=-0.47, two-tailed t-test p-value= 6.74e-08, n=122, 95% CI=(-0.6,-0.31). Data underlying the figure can be found in the Supplementary Data.