Statistics is integral to agricultural disciplines, yet many students face challenges in mastering it, particularly those with limited prior exposure. This study examines the impact of a problem-based learning (PBL) instructional design on student engagement in a graduate-level statistics course within an agricultural education context, leveraging learning analytics data for deeper insights. The study involved 31 students, with engagement measured using log data from a Learning Management System (LMS) and supplemented by written reflections and semi-structured interviews to capture experiences and perceptions.

Using machine learning techniques, students were categorized into high-performing (n = 13) and low-performing (n = 18) groups based on interaction patterns and academic performance. Results revealed significant differences in interaction with course content (p < .001), instructor (p < .001), system (p = .02), and overall engagement (p < .001). Low-performing students exhibited higher overall interaction but showed sharper declines in engagement post-exams and greater fluctuations across the semester. Conversely, high-performing students maintained consistent engagement, emphasizing the role of steady interaction in academic success.
Reflections and interviews highlighted that the PBL model effectively fostered student engagement and enriched learning experiences. The findings underscore the pivotal role of assessments, particularly exams, in driving engagement, while also revealing disparities in engagement patterns between performance groups.

This study demonstrates the potential of learning analytics to uncover actionable insights for improving instructional strategies. By understanding engagement trends, educators in agricultural education can design targeted interventions to better support diverse learner needs. The findings provide a roadmap for creating more effective and engaging learning environments across agricultural disciplines.