Figure A1.

(A) Relative error of differentiation estimates using T = 4608 samples, as a function of network size. The estimates of $\mathcal{D}$₁ get worse as network size increases, but the estimates for $\mathcal{D}$₂ stay accurate. (B) Relative error for networks with seven internal elements with increasing levels of measurement error. The estimate of $\mathcal{D}$₁ is greatly affected by even small amounts of noise, while $\mathcal{D}$₂ has only margin increase in error. (C) Minimum sample size to estimate differentiation for desired accuracy. The necessary sample size for $\mathcal{D}$₂ does not increase as the network size increases, however the sample size for $\mathcal{D}$₁ increases exponentially.