Longitudinal Modeling of Response Process Data

Most likely due to the challenges encountered interpreting time-dependent effects in a way that is familiar to most practitioners, eye tracking studies using longitudinal models remains quite rare. With the use of averaged, summed down or other aggregate versions of the original data, there is always a potential for the actual patterns of change to be obscured, severely limiting the breadth and depth of potential insights into time-varying processes.

​

We draw from the current developments in the field of educational assessment, supporting the use of item process data, such as eye movements, response times, and brain waves for validation and test development practices, and propose a longitudinal modeling approach, despite its popularity, has not been extended to modeling item-level eye-fixation data.

 

We explore the various ways in which item-level eye-tracking data can be analized to facilitate our understanding of skill development levels of examinees and to identify item characteristics that, if ignored, might undermine the validity of intended person-level inferences.

​

We will develop and test new and alternative latent growth and growth mixture model applications to estimate multiple-choice item processing curves, and to what extent construct-related, such as a correct response, or construct-unrelated factors, such as a change in the presentation order of the correct choice or brain activity explains some variance.

​

Below we present some material from our previous and current research projects.