Extended Data Fig. 5: IV-Bayes is a consistent estimator even with incorrect prior.
(a) (left) Example ground truth simulated effectome for linear simulations (off-diagonal weights drawn from uniform distribution ([0.1,0.2]), then on average 9- % are set to 0 and diagonal is set to 0.1) (a) (center) In ‘Correct prior’ condition prior mean is set to the true effectome. (a) (right) In ‘Incorrect prior’ condition prior mean is set to independent effectome (without diagonal). (b) Simulations of IV-Bayes estimator with correct (blue) and incorrect prior (orange) and the raw IV estimator (green) across number of samples (i.e., duration of recording) and resamplings of effectomes (mean ± s.d.; n = 10 simulations). Variance of prior mean has a constant (see Fig. 2c legend) of increasing size added to it (increasing left to right). R2 is measured between the estimate of the effectome and ground truth .
