Lecture by Conor Houghton:
Mutual information when there isn't so much data

In principle the best way to quantify the relationship between two signals is to use the mutual information; in practice, calculating mutual information seems to require more data than a typical electrophysiological experiment provides. Here, a simple Kozachenko–Leonenko estimator is presented which requires less data and may be useful in studying EEG signals; this should provide a new approach to quantifying the relationship, for example, between different sound stimuli and different ERPs.