P1-29: Assessing Student Conceptual Competencies using Bayesian Networks

By Wentao Yan, Dennis Pearl, and Matt Beckman, The Pennsylvania State University


Web applications are often used in labs or online assignments to help students explore statistical principles. Examining web log files allows evaluation of student interaction with these apps to judge their conceptual competency. This poster will demonstrate an education assessment system aimed at giving instructors or individual students a quantitative measurement of students’ competency in course concepts. We achieve this by building Bayesian Networks to model causal relationships between conceptual competency (θ), associated tasks (x) and attributes of the individual (a) in the assessment. Our models are fit using MCMC algorithms to make inferences about the posterior distributions of latent variables (θ) given observables. We applied our models on data for 75 introductory statistics students using interactive songs in project SMILES (www.causeweb.org/smiles), which ask students to respond to prompts on statistical concepts. We analyzed the interaction of students with those prompts to judge conceptual knowledge before listening to the song.

Poster Session - P1-29 - Assessing Student Conceptual Competencies using Bayesian Networks.pdf