Validation - SR Research EyeLink Portable Duo User Manual

The corneal reflection status error message "MISSING", highlighted in red,
indicates that the corneal reflection is not visible to the camera. See section 3.4
for details on how to set up corneal reflection properly.
All status flags remain on for a minimum of 200 milliseconds, even if the
condition that caused the warning or error to be raised lasts for less than 200
milliseconds.

3.8 Validation

It is important that any problems with the calibration be identified and
corrected before eye-movement recordings containing inaccurate and poor
quality data are collected. By running a validation immediately after each
calibration, the accuracy of the system in predicting gaze position from pupil
position can be established. If performance is poor, the calibration/validation
cycle should be repeated before data collection begins.
During validation, targets are again presented to the participant in a random
order, similar to the calibration procedure. When the participant fixates these
targets, the calibration model is used to estimate the gaze position of the
participant, and the error (difference between actual target position and
computed gaze position) is calculated. Note: a scaling factor is built in for
automatically-generated validation points to pull in the corner positions (see the
'validation_corner_scaling'
used to limit validation to the useful part of the display and test the calibration
accuracy on uncalibrated points.
The gaze-position error comes largely from errors in fixation data gathered
during the calibration/validation, which come from two sources: the eye-
tracking system and physiological eye-movement control. The EyeLink system
has extremely low eye-position noise and very high resolution, and corrects for
small head motion during calibration and recording. These common sources of
error in the eye-tracking system are virtually eliminated. One physiological
source of calibration inaccuracy is the natural variability of participant's ability
to accurately fixate the targets. Vergence eye movements also contribute – this
can be seen clearly during validation with binocular gaze position displayed.
For calibrations, it is possible that one or more targets may be fixated with an
error of 1° or greater. Poor eye/camera setup can cause a highly distorted
calibration pattern thus magnifying small errors. Some participants may show
substantial drifts in gaze position during fixations or may not fixate carefully,
adding to the errors.
60 An EyeLink Portable Duo Tutorial: Running an Experiment
© 2016-2017 SR Research Ltd.
command setting in the CALIBR.INI file). This is