A modern test theory approach to selecting eye tracking stimuli
No Thumbnail Available
Authors
Meeting name
Sponsors
Date
Journal Title
Format
Thesis
Subject
Abstract
Researchers who conduct eye-tracking studies often consider stimuli to be interchangeable without much consideration for item effects. For example, the saliency of distractor objects in an image plays a role in the difficulty of spotting a specific target object. This, in addition to using a large number of images, use up the resources of both the researcher and participant. The goal of this project is to explore the application of item response models to eye tracking data in order to reduce the number of images used in a study while keeping similar amounts of information. Specifically, I use data from an image memorability study by Bylinskii et al. (2015) to fit item response models formulated in a GLMM framework. Associated memorability scores are used as a standard of comparison for parameter estimates in the item response models. This method provides a way to select images of varying difficulty and to thin out images that overlap in the information they provide. Alternative link functions are explored for use with eye tracking data that is not dichotomous and simulations are conducted to assess various thinning methods as well as their stability. Overall, models tend to retain their predictive ability as the number of images are reduced. These findings suggest that researchers can decrease the number of images used in a study, given that they are high quality and cover a range of difficulty levels. This decrease then saves researchers and participants time and resources.
Table of Contents
DOI
PubMed ID
Degree
Ph. D.