Information hidden in signals and macromolecules I. Symbolic time-series analysis

We describe the conceptual background and practical implementation of some recently developed techniques for the analysis of symbol sequences and symbolic time series. We emphasize their associated software realization, the WinGramm suite of programs, that includes programs for the calculation of conditional entropies, context-free grammatical complexity, algorithmic distance and redundancy, as well as for the generation of surrogates that preserve symbol pairs and triplets. We demonstrate the usefulness of these programs by means of two illustrative examples, taken from computational neuroscience. In the first one, we obtain evidence of the Markovian character of the cortical inter spike intervals of the rat before penicillin treatment, and its disappearance afterwards. In the second one, we extend previous investigations about neural spike-trains generated by the isolated neuron of the slowly adapting stretch receptor organ (SAO), in order to classify sequences of different length of known neural behaviors. We include new spike trains, digitized employing the optimal partition procedure described by Steuer, Molgedey, Ebeling, & Jimenez-Montaño, (2001).