Closed samiamseid closed 3 months ago
My understanding has been that these artifacts are expected and excluded from downstream analysis. If this assumption is incorrect, then instead of excluding these we should add a tag that fails this data, and we can have a larger discussion about fixing it
@samiamseid I had not heard that these artifacts are expected and filtered out of analysis. Can you provide more information or tag folks who have more information?
this assumption is simply by virtue of the fact that there has always been a metric called "wheel artifacts (>100cm/sec)", but there is nothing in the QC SOP about a flag or fail threshold for that metric. These wheel artifacts are not that uncommon and in fact the QC SOP example for "good" wheel encoder plots has 6 artifacts
I recognize this isn't a principled assumption, just a statement of how things have always been done, and to my knowledge nobody downstream has ever commented on the data suggesting that a mouse just teleported 10 meters
@seanmcculloch seems like this is still an issue, is that correct?
Set Y axis limit based on everything besides first frame.
Should be done with this by EOW
A fix has been implemented and is in review now https://github.com/AllenInstitute/mindscope_qc_metrics/pull/41
Completed and merged into prod.
Metric / Plot Impacted Wheel Encoder Plots
Update Motivation Sometimes at the beginning of a wheel encoder timeseries, there are artifacts which suggest motion which did not actually occur. Due to the way the plots are generated, these artifacts stretch the y axis of the plots to the point where the graph is no longer helpful
Example: http://mouse-seeks/qc/ophys/session/1287539392
Requested Feature/Solution Excluding the artifacts from the plots, or perhaps just using the 99th percentile for the y axis limits would allow the scale of the plot to remain useful even when there are artifacts in the encoder data.