Behavioral-only metrics from an instrumented online textbook support early prediction of final course grades in flipped statistics courses¶

Claim · All claims
Evidence · 1 study (2 entries) · associational r1 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Models using behavioral-only metrics engineered from students' textbook interactions were developed to detect at-risk students early in flipped introductory statistics courses. → Alice Xu 2026 q2 i? Eleven chapter-level features (six reading-related, five formative-assessment-related) were constructed per chapter, yielding 132 features across 12 chapters. → Alice Xu 2026 (2)

Evidence¶

Alice Xu 2026¶

Alice Xu, Icy (Yunyi) Zhang, Adam B. Blake and James W. Stigler. (2026). Modelability as a Strategy for Improving the Generalizability and Scalability of Predictive Models. Journal of Learning Analytics, 13(1). https://doi.org/10.18608/jla.2026.9099

q2 · i? · associational · r1

Case study modeling goal (RQ1): predicting final grade category early in the term from behavioral-only metrics in a flipped classroom setting. Data came from two courses (183 and 60 completing students) at two public research universities; no effect size is printed.

"we developed a model detecting at -risk students early in the courses , using behavioral -only metrics derived from students' uses of an online textbook in an introductory-level statistics course delivered in a flipped classroom setting."

Alice Xu 2026 (2)¶

Alice Xu, Icy (Yunyi) Zhang, Adam B. Blake and James W. Stigler. (2026). Modelability as a Strategy for Improving the Generalizability and Scalability of Predictive Models. Journal of Learning Analytics, 13(1). https://doi.org/10.18608/jla.2026.9099

q2 · i? · associational · r?

Feature engineering: reading sessions derived from enGauge timestamps yielded six features per chapter (e.g., session count, engagement ratio), plus five formative-assessment features; "In total, the model includes 132 features, with 11 features for each of the 12 chapters."

"Specifically, we constructed 11 chapter-level features: six related to reading behavior and five related to formative assessment performance."

Discussion¶

Related Claims¶

• Attendance, behavior, and course performance (the ABC indicators) reliably predict which secondary students are at risk of dropping out — the general claim this finding bears on, with the studies behind it
• Predictive models built in a modelable ecosystem generalize across institutions — related
• In-class chat engagement significantly predicts higher test scores in MAIC courses — related
• Attendance and online platform participation remained predictive of students' course grades during remote learning, with relationships as large or larger than pre-pandemic — related