Detecting best lag of embedding for modeling spike-wave discharges from experimental data

Abstract

Purpose. Optimal value of the embedding lag calculation is made. Lag is one of empirical parameters of mathematical models, used in autoregressive models for prediction, coupling analysis, signal classification etc. Methods. The first minimum in the dependence of the mutual information function on the time lag was detected. Results. The calculation showed that the optimal lag is about 8 sampling intervals (1/64 s or 1/8 of the characteristic oscillation period for the absence seizures). Discussion. The optimal lag is about 1/8 of the characteristic oscillation period was obtained for both epileptiform and background activity, including preictal and different stages of ictal activity, i. e. this time scale is present in the signal throughout the observation time.