So I’ve noticed there are nan values in the LFP in the neuropixels dataset for visual behaviour. Just curious why this would be and how to manage them.
Here is where I noticed them:
Session id: 715093703
probe id 810755801
channel: 850261534
I am upsampling the signal from 1250 Hz to 1500 hz as well to use a filter designed by another lab that’s superior to what I have. I’m concerned about what this will do to the signal.
Hi @AngCamp – this is due to brief intervals in which the connection to the probe was dropped, and no data was collected. The exact times and probes affected can be seen by querying session.invalid_times.
For that probe and session, there are 7 invalid intervals. These will appear as NaN values in the LFP data, and there will be no spike times associated with these epochs. This is uncommonly high, as most sessions/probes have all of the data intact.
The simplest fix is to just ignore that particular probe and analyze other sessions instead. But if that’s not an option, you can filter each of the continuous (non-NaN segments) separately, using the invalid_intervals table as a guide for where to split the data.
I was considering just dropping the probe but I wasn’t sure how much data I would be throwing out if I did that. If you say its uncommon I will just discard the probe.
Yeah, that’s what I’d recommend.