Tobii eye tracking missing data


Tobii eye tracking missing data

Author
Message
Willem
Willem
Partner Member
Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)

Group: Forum Members
Posts: 17, Visits: 24
I noticed in the eyetracker output that there are very few recording including missing data. This is odd, since missing data is to be expected in eye tracker data due to blinks.

It appears that when the eyetracker does not detect an eye (e.g., during a blink), it also does not record any data.

As a consequence, it is difficult to figure out how much missing data there is, or to use methods such as linear interpolation to fill up gaps in the data.

My question is whether this is something that can be fixed. Is there a way to use the tobii plugin to get recording of ALL measurements? I know that with Tobii Studio or with the E-prime plugin for Tobii eye trackers, it is possible to also get rows containing missing data.

Best,
Willem
Willem
Willem
Partner Member
Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)

Group: Forum Members
Posts: 17, Visits: 24
Willem - Tuesday, June 19, 2018
I noticed in the eyetracker output that there are very few recording including missing data. This is odd, since missing data is to be expected in eye tracker data due to blinks.

It appears that when the eyetracker does not detect an eye (e.g., during a blink), it also does not record any data.

As a consequence, it is difficult to figure out how much missing data there is, or to use methods such as linear interpolation to fill up gaps in the data.

My question is whether this is something that can be fixed. Is there a way to use the tobii plugin to get recording of ALL measurements? I know that with Tobii Studio or with the E-prime plugin for Tobii eye trackers, it is possible to also get rows containing missing data.

Best,
Willem

I've been thinking about it some more and I also fear an alternative limitation.

I use markers to indicate when what happens in the eye tracker data. For example:

/ stimulustimes = [0 = gray, marker_gray; 2000 = picture_negative_order2, marker_picture; 8000 = white, marker_white])


I then use these markers to figure out the starting point of each trial. In this case, I can say that the first observation of the marker_gray marker is the moment when the trial begins. But what if someone blinked at that time? Does this mean that the first time the marker_gray marker appears is actually the first time an eye is observed in that specific trial? This is not the same as when the trial actually began.

Is this an issue?




Dave
Dave
Supreme Being
Supreme Being (428K reputation)Supreme Being (428K reputation)Supreme Being (428K reputation)Supreme Being (428K reputation)Supreme Being (428K reputation)Supreme Being (428K reputation)Supreme Being (428K reputation)Supreme Being (428K reputation)Supreme Being (428K reputation)

Group: Administrators
Posts: 8.9K, Visits: 38K
Willem - Wednesday, June 20, 2018
Willem - Tuesday, June 19, 2018
I noticed in the eyetracker output that there are very few recording including missing data. This is odd, since missing data is to be expected in eye tracker data due to blinks.

It appears that when the eyetracker does not detect an eye (e.g., during a blink), it also does not record any data.

As a consequence, it is difficult to figure out how much missing data there is, or to use methods such as linear interpolation to fill up gaps in the data.

My question is whether this is something that can be fixed. Is there a way to use the tobii plugin to get recording of ALL measurements? I know that with Tobii Studio or with the E-prime plugin for Tobii eye trackers, it is possible to also get rows containing missing data.

Best,
Willem

I've been thinking about it some more and I also fear an alternative limitation.

I use markers to indicate when what happens in the eye tracker data. For example:

/ stimulustimes = [0 = gray, marker_gray; 2000 = picture_negative_order2, marker_picture; 8000 = white, marker_white])


I then use these markers to figure out the starting point of each trial. In this case, I can say that the first observation of the marker_gray marker is the moment when the trial begins. But what if someone blinked at that time? Does this mean that the first time the marker_gray marker appears is actually the first time an eye is observed in that specific trial? This is not the same as when the trial actually began.

Is this an issue?




As far as I understand it, this shouldn't be the case. The marker should appear in the data stream when it was sent, regardless of whether there was a blink at that time or not. Does your data suggest otherwise?

Willem
Willem
Partner Member
Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)

Group: Forum Members
Posts: 17, Visits: 24
Dave - Wednesday, June 20, 2018
Willem - Wednesday, June 20, 2018
Willem - Tuesday, June 19, 2018
I noticed in the eyetracker output that there are very few recording including missing data. This is odd, since missing data is to be expected in eye tracker data due to blinks.

It appears that when the eyetracker does not detect an eye (e.g., during a blink), it also does not record any data.

As a consequence, it is difficult to figure out how much missing data there is, or to use methods such as linear interpolation to fill up gaps in the data.

My question is whether this is something that can be fixed. Is there a way to use the tobii plugin to get recording of ALL measurements? I know that with Tobii Studio or with the E-prime plugin for Tobii eye trackers, it is possible to also get rows containing missing data.

Best,
Willem

I've been thinking about it some more and I also fear an alternative limitation.

I use markers to indicate when what happens in the eye tracker data. For example:

/ stimulustimes = [0 = gray, marker_gray; 2000 = picture_negative_order2, marker_picture; 8000 = white, marker_white])


I then use these markers to figure out the starting point of each trial. In this case, I can say that the first observation of the marker_gray marker is the moment when the trial begins. But what if someone blinked at that time? Does this mean that the first time the marker_gray marker appears is actually the first time an eye is observed in that specific trial? This is not the same as when the trial actually began.

Is this an issue?




As far as I understand it, this shouldn't be the case. The marker should appear in the data stream when it was sent, regardless of whether there was a blink at that time or not. Does your data suggest otherwise?

Thank you for the response Dave.

There are some reasons I do suspect otherwise. For instance, although it is rare, some of the markers that should be there are missing in my eye tracker data. In my specific case, I have 3 markers for each trial, indicating the onset of a gray picture, a photograph, and a white picture. For one of my participants, the eye tracker data stream goes from a white picture marker in trial 66 immediately to a white picture marker of trial 67, thus skipping the gray and photograph picture markers. 

Additionally, if I calculate the time difference between the first gray marker and the first photograph marker for each trial (which should be 2000ms), I get the following results:


Min. 1st Qu. Median  Mean 3rd Qu.  Max.  NA's  
250  2000     2000    2036 2000     4467  4


So sometimes the difference between the onset of the first gray marker and the onset of the photograph marker is only 250ms, rather than the 2000ms it should be.
seandr
seandr
Supreme Being
Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)

Group: Administrators
Posts: 1.2K, Visits: 916
Hello,
I've checked the code and we do indeed discard gaze point samples associated with "no validity" (e.g., an eye blink).
This is something we could change (or make configurable), although unfortunately not for version 4. If we fixed this in version 5, would you be able to upgrade to take advantage of it?
-Sean


Willem
Willem
Partner Member
Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)

Group: Forum Members
Posts: 17, Visits: 24
seandr - Wednesday, June 20, 2018
Hello,
I've checked the code and we do indeed discard gaze point samples associated with "no validity" (e.g., an eye blink).
This is something we could change (or make configurable), although unfortunately not for version 4. If we fixed this in version 5, would you be able to upgrade to take advantage of it?
-Sean


Thanks Sean; that explains the results.

We currently have an Inquisit Departmental Licenses plan for Inquisit 4 and 2 Tobii Plugin licenses (not sure if these are tied to Inquisit 4). So in order to make use of the fix we would have to upgrade 2 computers to Inquisit 5. Depending on the costs, I may be able to persuade our department head to do this.
seandr
seandr
Supreme Being
Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)Supreme Being (64K reputation)

Group: Administrators
Posts: 1.2K, Visits: 916
Is it OK for me contact you at the email address you've registered with this forum to discuss options? Or you can send a quick note to sales@millisecond.com if you'd prefer to use a different email.

Thanks,
Sean
Willem
Willem
Partner Member
Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)Partner Member (659 reputation)

Group: Forum Members
Posts: 17, Visits: 24
seandr - Wednesday, June 20, 2018
Is it OK for me contact you at the email address you've registered with this forum to discuss options? Or you can send a quick note to sales@millisecond.com if you'd prefer to use a different email.

Thanks,
Sean

That's OK; thanks again.
GO


Reading This Topic


Login
Existing Account
Email Address:


Password:


Social Logins

Select a Forum....






Millisecond Forums


Search