Bayesian Hierarchical Models for Student Outcomes in Cyber-Enabled Science Classrooms

• Activity Number: 686
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Yuanzhi Li* and Daniel Coster
• Companies: and Utah State University
• Keywords: Bayesian hierarchical model ; Cyber-enabled learning ; MCMC

Abstract

Project Purpose: This paper is based on the Discovery Research K-12 NSF-funded project, "Cyber-enabled learning: Digital Natives in Integrated Scientific Inquiry Classrooms". Our goal was to determine whether teacher professional development (PD) utilizing cyber-enabled resources lead to meaningful student learning outcomes. We use 8th grade student end-of-year scores for students with teachers who underwent PD.

Methodology: A Bayesian hierarchical model is used. Students are grouped by socio-economic status (low or high household income). Teachers' performances, assessed at two time points, are latent random effects in the model, and are modeled with a multivariate normal prior having means as the linear combination of external skills instruments. Using Markov Chain Monte Carlo (MCMC) algorithms, we make posterior inferences from summaries of the MCMC samples.

Results: Teachers' performances at the beginning of year and at the end of year have positive effects on students' CRT scores with a high correlation around 0.83 between pre- and post- teacher performances. Certain instrument scales had a greater influence on teachers' performances than others.