thomaskreuz
commented
1 year ago Dear Ruairi,
sorry for the very late reply. We just released a new version of PySpike (0.8.0) where we also included the rate-independent SPIKE-distance. I would recommend to try this (and compare it against the regular ISI-and SPIKE-distance, without and with shuffled ISI). Maybe it could help you understand what is going on in your example.
For more information please have a look at these two papers:
Satuvuori E, Mulansky M, Bozanic N, Malvestio I, Zeldenrust F, Lenk K, and Kreuz T: Measures of spike train synchrony for data with multiple time-scales J Neurosci Methods 287, 25 (2017) PDF: https://drive.google.com/file/d/1xOepp9WN0luODH9qF6i5vVRimqC9WweA/view
Satuvuori E, Kreuz T, Which spike train distance is most suitable for distinguishing rate and temporal coding? J Neurosci Methods 299, 22 (2018) PDF: https://drive.google.com/file/d/1tLAWXEBkIXBbw02aD5reFjTLnvlGFJEx/view
Hello,
Thanks for uploading this.
I notice that the measures in this repo find relationships between spike trains even when ISIs are shuffled or jittered. For example, here is a spike train syncrhony measure applied to a set of simultaneously recorded spiketrains. There is clear structure that is retained when the inter spike intervals are shuffled.
Is this the expected behaviour? If so, how do you propose one uses these measures to find relationships not related to spike rate etc? In the above I compared the observed to bootstrap samples of shuffled ISIs but this doesnt seem like the best way to go.
Many thanks, Ruairi