P2-29: Tips and Pitfalls of Converting to Simulation-based Inference for Large Lectures

Information

This poster details our experience converting a large-enrollment introductory statistics course from a traditional curriculum to simulation-based inference curriculum. Typical enrollment for introductory statistics courses at Penn State is approximately 2000 students per semester. Students attend large lectures of size 200-300 twice per week. After each lecture, students participate in an active-learning lab in smaller groups of approximately 80. During the spring semester of 2017, a single large-enrollment classroom of 240 students was chosen to pilot the new curriculum using simulation-based inference and assessed using high-frequency, low-stakes exams. We used the Lock⁵ textbook in its second edition, extensively sourcing activities, lectures, and examples from textbook-provided instructor materials. A control class was held the same semester at a similar timeslot (8 am lectures) with an identical exam schedule. We compare learning outcomes using a common cumulative exam and responses to the GOALS instrument:  a series of questions designed at the University of Minnesota to measure statistical reasoning in a first course in statistics. Because the courses are currently underway, results will not be available until USCOTS.

The large-enrollment format of the classroom presented several challenges. Lectures are large enough to necessitate the almost exclusive use of projected slides. Real-time illustrations were performed using Doceri, an app that allows annotation of slides and any other screen. It also allows the instructor to move around the classroom while controlling the podium remotely. To encourage students’ active interaction with materials we relied heavily on active learning labs. These take place in a computer lab under the supervision of a graduate teaching assistant and an undergraduate learning assistant. Students work in small groups and submit their work through an online quiz based on the lab activity. To encourage student engagement we used data-generating experiments such as coin-flipping, marble-drawing, and heart-rate measuring. Many, but not all, of these experiments were sourced from the Lock^5 text. Students used Google Forms to collect responses in real-time, and used this self-generated data in Minitab Express or StatKey to perform analysis. Challenges addressed in this poster include meaningful assessment of student work during labs, preparation and delivery of online exams through course management sites, and the protection of student-sourced real-time data from sabotage by anonymous students.