Estimation of single-neuron model parameters from spike train data

Estimating parameters for models of neurons requires a quantitative comparison between the model output and empirical data. The present study compares three error functions: voltage time-series (VTS), cumulative voltage integrals (CVI), and phase histograms (PH). In two test cases, predefined models were used to produce target data and to compare the efficacy of the three error functions when they were used to recover the target data. In a third example, empirical data were used to parameterize a model. VTS was found to be inferior, whereas as CVI and PH were similar and effective. Reliable parameters were derived from analyzing as few as two data sets. r 2004 Elsevier B.V. All rights reserved.