NONPARAMETRIC ESTIMATION AND NUMERICAL EVALUATION OF TIME POINT OF OCCURRENCE OF EVOKED POTENTIAL IN MAGNETOENCEPHALOGRAM

In order to investigate the rate of transferring information from the sensory organ to the brain hemisphere through the averaged magnetoencephalogram (MEG) data, we propose an estimator of the time point at which cerebral evoked potential (CEP) occurs. We regard the averaged MEG as a stationary process before the occurrence of the CEP and as a nonstationary process with a mean function, whose functional form is unknown, after the occurrence of the CEP. We approximate the mean function in a neighborhood of the time point of the occurrence by a linear function. We propose an estimator of the time point of the occurrence, that is, the first time point at which the slope of a linear function becomes nonzero. We compare, by simulation studies, the proposed estimator with some other estimators. We find that our estimator is reasonable when the mean value gradually changes. Finally, we show an example of application of the proposed method to real MEG data.