Analyzing the Effectiveness of Design Elements Within Statistical Teaching Apps Through Bayesian Network Analysis

open_access
Open Access
Author:
Pechulis, Nathan
Area of Honors:
Statistics
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
Neil J Hatfield, Thesis Supervisor
Matthew D Beckman, Thesis Honors Advisor
Keywords:
Bayesian networks
behavioral data analysis
educational technology
user engagement
Bayesian inference
latent class models
Abstract:
The Book Of Apps for Statistics Teaching is a resource created for undergraduate students to enhance their statistics knowledge through free, interactive apps created via the Shiny package on R. All interactions of app users since 2020, from simply opening the app to completing an activity, has been collected into a dataset of log-files. This data provides a rich foundation for analyzing patterns in student learning and engagement with e-learning applications. This study explores how these log-files can be transformed into actionable insights about student e-learning via Bayesian networks and the rjags package on R. By constructing a probabilistic framework like this, we can capture relationships between observed variables and unobserved, underlying engagement levels, thus improving our ability to understand how students interact with certain educational tools. The analysis ultimately revealed that Interactive Challenges and Right/Wrong-style feedback show a positive association with user performance/engagement under this framework.