Fig 1. Data misfit surface perturbed by experimental and model error.

(a) Membrane voltage, V_exp(t_i), recorded in discrete time t_i, i = 0…n (cross symbols); useful membrane voltage, V_use(t), obtained from an ideal measurement apparatus (black line); membrane voltage state variable of the conductance model, V_mod(t) (red line). Experimental error: ϵ_exp(t_i) = V_exp(t_i) − V_use(t_i). Model error: ϵ_mod(t) = V_mod(t) − V_use(t). (b) Lines of constant data misfit, δc = f(σ, ζ), about the global minimum $p_0^{*}$. Different noise realizations, ζ₁ (ζ₂), shift the global minimum ${\mathit{p}}_0^{*}\to {\mathit{p}}_{\sigma {\zeta }_1}^{*}$ (${\mathit{p}}_0^{*}\to {\mathit{p}}_{\sigma {\zeta }_2}^{*}$). Noise also tilts the principal axes of the data misfit ellipsoid (red/blue arrows) and modifies the principal semi-axes (λ_i,λ_j). (c) RVLM neuron model membrane voltage V_exp (black line) induced by current injection I_inj (blue line). Additive noise ϵ_σζ is incorporated in the model data. (d) Posterior distribution function π(p_k) of parameter p_k, k = 1…K.