Table 1.

Parameter Priors for model parameters including the observation model, neuronal sources, and experimental effects

Parameter ϑi = πiexp(Θi)	Interpretation	Prior
		Mean: πi	Variance: Θi = N(0,σi)
Observation model
αu	Exogenous white input	παu = 0	σαu = 1/16
αs	Channel specific white noise	παs = 0	σαs = 1/16
αc	White noise common to all channels	παc = 0	σαc = 1/16
βu	Exogenous pink input	πβs = 0	σβu = 1/16
βs	Channel specific pink noise	πβc = 0	σβs = 1/16
βc	Pink noise common to all channels	πθi = 1	σθi = exp(8)
θ1…s	Lead-field gain	πλ = 0	σλ = 1
λ	Noise hyperparameter
Neuronal sources
κe/i	Excitatory/inhibitory rate constants	πκe = 4 ms− 1πκi = 16 ms− 1	σκe = 1/8 σκi = 1/8
He/i	Excitatory/inhibitory maximum post-synaptic potentials	πHe = 8 mV πHi = 32 mV	σHe = 1/16 σHi = 1/16
γ1,2,3,4,5	Intrinsic connections	πγ1 = 128	σγ1 = 0
πγ2 = 128	σγ2 = 0
πγ3 = 64	σγ3 = 0
πγ4 = 64	σγ4 = 0
πγ5 = 4	σγ5 = 0
AF	Forward extrinsic connections	πAF = 32	σAF = 1/2
AB	Backward extrinsic connections	πAB = 16	σAB = 1/2
AL	Lateral extrinsic connections	πAL = 4	σAL = 1/2
C	Exogenous input	πC = 1	σc = 1/32
di	Intrinsic delays	πdi = 2	σdi = 1/16
de	Extrinsic delays	πde = 10	σde = 1/32
Design βki	Trial specific changes	πβki = 1	σβki = 1/2

In practice, the non-negative parameters of this model are given log-normal priors, by assuming a Gaussian density on a scale parameter, Θ_i = N (0,σ_i), where ϑ_i = π_iexp(Θ_i), and π_i is the prior expectation and σ_i² is its log-normal dispersion.