Skip to main content
. 2018 May 23;38(21):4859–4869. doi: 10.1523/JNEUROSCI.3440-17.2018

Table 1.

Parameter values used in grid search and final fit values

Parameter (unit) Grid range Grid steps (spacing) Linear decay Exp decay Const drift Var drift
γ (s−1) (23, 29) 25 (log) 95.2 ± 15.2 98.7 ± 15.9 99.5 ± 19.1 94.2 ± 18.0
m (22, 27) 11 (log) 7.1 ± 0.74 7.0 ± 0.73 7.3 ± 0.95 7.0 ± 0.96
κ (23, 27) 9 (log) 22.0 ± 2.5 22.2 ± 2.5 26.0 ± 3.5 31.9 ± 5.10
τ (s) (2−6, 2) 8 (log) 0.18 ± 0.093 0.18 ± 0.10 0.14 ± 0.044 0.44 ± 0.17
ν (s−1) (1, 15) 15 (lin) 9.2 ± 0.70 9.3 ± 0.71 9.2 ± 0.70 9.3 ± 0.72
o (s) (0.1, 0.3) 11 (lin) 0.21 ± 0.008 0.21 ± 0.008 0.21 ± 0.008 0.21 ± 0.008
ω (s−2) (2−2, 24); 0 13 (log) + 1 0.16 ± 0.13
λ (s−1) (0, ln(2)) 9 (log) 0.017 ± 0.013
χ0 (s−1) (27, 214); ∞ 15 (log) + 1 3415 ± 1082 4637 ± 1484

The parameters are gain of spiking activity γ, spike threshold m, neural tuning curve width κ, spike survival rate τ, rate parameter ν, and offset o for the exponential component of the latency distribution, decay rate ω (linear decay model) or λ [exponential (Exp) decay model], and drift rate χ0 (drift models). For each parameter we report the range over which it was varied in the grid search (with an additional discrete value reflecting no delay effects for ω and χ0), the number of steps in the grid search and their spacing within the given range (logarithmic or linear), and the maximum likelihood estimates (mean over participants ± SE) for each of the four model variants. Threshold values m are always rounded to the nearest integer. The fit values for χ0 exclude one outlier in the constant (Const) drift variant (with χ0 = 8.3 × 107 s−1), and one outlier in the variable (Var) drift variant (with χ0 = 2.4 × 106 s−1), reflecting individual model fits with virtually no drift.