Skip to main content
Open Access Publications from the University of California


UCLA Electronic Theses and Dissertations bannerUCLA

Modeling cognitive control: Using cross-classified IRT and structural equation modeling to understand brain and behavior predictors of academic achievement


Cognitive control skills are foundational to goal-directed action and related to children’s academic achievement. These skills are often quantified using observed measures of children’s brain (e.g., ERN, CRN, Pe) and behavior (e.g., accuracy, reaction time) on computerized tasks. While computerized tasks afford precision in measurement, observed measures of behavior do not allow for examination of differences in performance between trials that may result from varying task features (e.g., trial difficulty, trial randomization) or children’s interactions with the task (e.g., previous trial performance impacting next trial performance). In this investigation, I leverage item response theory (IRT) models to identify task features that impact children’s behavioral performance as well as accommodate for inter-trial variation in Go/No-Go performance among children. Results indicate that task features, such as trial difficulty, impact children’s likelihood of making an error and also impact reaction time performance. Further, studies utilizing tasks like the Go/No-Go typically report linear regressions between an academic outcome and a single score of behavioral performance. However, relations between individual scores and real-world skills, including academics, are relatively weak and inconsistent across studies. In this project, I compare three approaches for modeling relations between cognitive control and academic skills – a traditional regression approach, an observed measure SEM model, and an adjusted (predicted) measure SEM model. Nine behavioral and four ERP measures of Go/No-Go task performance are assessed as indicator variables for a cognitive control construct. Findings show that SEM models that incorporate multiple measures of cognitive control are superior for predicting variance in academic outcome measures as compared to regression approaches. Overall, this study emphasizes that multiple measures from one computerized assessment of cognitive control can be leveraged to predict moderate variance in academic skills for children in early elementary school.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View