HI all, thanks for all your work on this project!
I'm trying to make sense of my results with CASCADE using model Global_EXC_7.5Hz_smoothing200ms
My data is 2p recorded from interneurons in the spinal cord at 8.46hz using Gcamp6s. If it matters, the recordings are performed ex vivo at 28-30c rather than at 37c so I might expect slightly slower kinetics. I initially used the google collab notebook, and have now installed locally and have found a similar problem.
I have noticed that while the input DFF and the output spike_prob match in the number of frames (including the nans as 'frames', i.e. just considering them zeros for the purpose of aligning against the DFF) the peaks do not line up with the stimuli nor with the original DFF anymore. The total amounts is ~5frames (~600ms) which doesn't seem to drift much throughout the recording, e.g. the two examples here are from the start and near the end of the recording.
I also tried using the 3hz spinal cord model, but those results looked quite wacky, which is I assume due to the dramatic difference in frame rates. I have also tried the causal kernel, which does shift things a bit in the right direction, but doesn't actually fix the underlying issue. I've tried it with multiple data sets and I keep running into the same issue. I am seeing 32 nans on either end, which I think is correct, and the number of frames line up correctly so I'm just struggling to figure out if there is a simple explanation that I'm missing. Thanks!
HI all, thanks for all your work on this project!
I'm trying to make sense of my results with CASCADE using model Global_EXC_7.5Hz_smoothing200ms
My data is 2p recorded from interneurons in the spinal cord at 8.46hz using Gcamp6s. If it matters, the recordings are performed ex vivo at 28-30c rather than at 37c so I might expect slightly slower kinetics. I initially used the google collab notebook, and have now installed locally and have found a similar problem.
I have noticed that while the input DFF and the output spike_prob match in the number of frames (including the nans as 'frames', i.e. just considering them zeros for the purpose of aligning against the DFF) the peaks do not line up with the stimuli nor with the original DFF anymore. The total amounts is ~5frames (~600ms) which doesn't seem to drift much throughout the recording, e.g. the two examples here are from the start and near the end of the recording.
I also tried using the 3hz spinal cord model, but those results looked quite wacky, which is I assume due to the dramatic difference in frame rates. I have also tried the causal kernel, which does shift things a bit in the right direction, but doesn't actually fix the underlying issue. I've tried it with multiple data sets and I keep running into the same issue. I am seeing 32 nans on either end, which I think is correct, and the number of frames line up correctly so I'm just struggling to figure out if there is a simple explanation that I'm missing. Thanks!