Closed thartbm closed 5 years ago
Fixed the way angles are calculated, made the window-based outlier removal a bit more flexible, and also found an error in how target and home position are stored for no-cursor and error-clamp trials (they still used the old, hard-coded home position). This should all combine to prevent outlier weirdness. If not, just re-open the issue.
The usual strategy is to rotate the trajectory data so that the target is at 0 degrees, and then get the discrepancy at some point using arctan2. This means that values should be between -180 and 180 degrees. However, we get values in the range of 300+ degrees, and this seems to be related to the windowed outlier removal. Need to look into this.