Efficacy of Normalized Compression Similarity (NCS) to detect long-range spike interactions.

<p>(A) NCS, Pearson coefficient exerted on 100 couples of independent uniformly distributed binary sequences (1000 bits). Both functions do not show bias. (B) To test the capacity of NCS to detect significant interactions within time windows of 1000 ms, we fixed in the center of the first time-window a binary random pattern (100 ms long, first plot above). The same pattern was replicated in a sequence of 47 time-windows drifting it from the initial to the window end. NCS, Pearson coefficient and the time lagged cross-correlation (the maximum value) were evaluated on these sequences: NCS is able to detect the interaction along the entire drifting process while Pearson coefficient is able to return significance only when the reference pattern is almost aligned in both sequences (drifts number 23–27). The time lagged cross-correlation detects weak significances in the data because of the spikes that not belong to the pattern.