# Model Eye Tracking Data with Stochastic Processes

This is a general question and I am asking for some creative solutions/tips to model as I am not very well versed in this domain. I am looking at eye tracking data and you can easily see that for each of the participants there is a "flow/gradient" of the eye tracking data through regions of interest with respect to time. How would I model this as a stochastic process? Does anybody have some stuff where I can catch up on?

To put it in more concrete terms I would like to give an example: One eye tracking point is a 3-dim. vector $(x_i, y_i, t_i)$, where $x_i, y_i$ represent the coordinate axis and $t_i$ is the time domain. My simplest approach is to build a Markov chain and compute the invariant distribution (with some adjustments s.t. the invariant distribution exists and can be approximated well). However this would throw away the time domain. Are there any approaches where I could model the data including the time domain to receive something like a stochastic process?

Br, Carl

• okay i will try to do this by modelling markov processes, it shouldnt be too hard. just read chapters 4 and 5 of spooks: introduction to markov processes – Carl Aug 31 '17 at 21:18