search

VPNLirvine / trEYEcopa

Code for eyetracking the TriCOPA videos
2 stars 1 forks source link

Reject individual trials based on blinks #51

Open brandonhackney opened 6 months ago

brandonhackney commented 6 months ago

At least one subject was falling asleep on some trials, and another was blinking like 5 times a second throughout the experiment. The data from trials like that will be unusable, but we don't want to throw the entire subject out in case some good trials remain. So implement some sort of trial-level blink detection and set a threshold where, say, if your eyes are closed for >30% of the video duration, we just reject the entire trial and return a NaN or something in selectMetric.

brandonhackney commented 4 months ago

There is code in the pupil branch that detects blinks and replaces them with interpolated values instead of just censoring them - use that as a starting point for at least finding the timepoints impacted by blinking. Whether you still want to interpolate or just censor is a matter of debate.