Likelihood Methods for Neural Spike Train Data Analysis - Computational Neuroscience: A Comprehensive Approach - page 266

1

Exponential
Gamma
Inverse Gaussian

0.5

0

0 0.5 1

Model Quantiles

Figure 9.4
Kolmogorov-Smirnov plots for the fits of the exponential (dotted line), gamma (solid
line), and inverse Gaussian (solid bold line) models to the neural spike train in Figure
9.2. The parallel diagonal lines are the 95% confidence bounds for the degree of
agreement between the models and the spike train data. By this criterion, statistically
acceptable agreement between a model and the data would be seen if the KS plot for
that model fell entirely within the confidence bounds.

Exponential Gamma Inverse Gaussian

ˆ l ˆ l ˆ l

ˆa ˆm

q

0.0325 0.805 0.262 30.76 12.1

CI [0.0283 0.0367] [0.678 0.931] [0.208 0.316] [24.46 37.06] [9.9 14.3]

ISI 30.77 ± 30.77 30.73 ± 34.74 30.76 ± 49.0

AIC 8598 8567 8174

KS 0.233 0.2171 0.1063

Ta b l e 1 : The row is the maximum likelihood estimate q. CI (95% confidence inter-
val for the parameter); ISI (interspike interval mean and SD); AIC (Akaike's Infor-
mation Criterion); KS (Kolmogorov-Smirnov statistic).