Flipping a Statistical Analysis Course

I wanted to share my reflections on flipping a course in Fall 2015 with my colleague, Dan Naiman, Professor of Applied Mathematics and Statistics at Johns Hopkins University. The course is 550.111: Statistical Analysis I. Previously, this 4-credit course met four time per week for 50 minutes – three lectures by faculty and one small-group meeting led by a Teaching Assistant (TA).

Text reading flipping the classroom with the classroom upside downStarting in Fall 2015, students watched several short videos (anywhere from 5 to around 20 minutes each) before the week started. Students then met once for a 75-minute lecture with the instructor and twice in small-groups with a TA. During these sessions students solved problems in teams of three with a TA available for help as needed.

In Fall 2016, we amended the format slightly: students met in a large lecture twice a week, on Mondays and Fridays, and met in discussion sections twice a week, on Tuesdays and Thursdays. This was in response to feedback from students indicating that they preferred a bit of additional face-to-face meeting time with the instructor. The Monday-Friday lecture times also made homework submission and certain aspects of course planning (such as exams) easier to handle.

We made this change because we wanted students to spend more time in small groups solving problems and engaged in activities, as opposed to simply listening to a lecture.

What did we learn? I would strongly advise those interested in flipping a class to keep the videos short. They should be about five minutes each. This allows each video to cover a discrete topic, and it’s about as long as students will watch in one session. Recording shorter videos is easier on the instructor as well. The video production took longer than I expected. For each video, Dan and I would first construct a slideshow, and then we would record it using the software program Camtasia. My colleague, Dan, did an excellent job with video production, and we generated significant video content before the start of the semester. I would also advise instructors to complete all video production before the start of the semester; we still had a few videos to produce during the semester, and this was a challenge. I found I was pressed to finish those additional videos in time. We plan to revisit, edit, and potentially add more videos before the next course offering. Specifically, we are considering animations and possible hand-written solutions.

We conducted clicker quizzes at the beginning of each lecture to motivate students to watch the videos. However, based on the video logs, quiz results, and the questions they asked, I found a number of students were not fully prepared. Their questions were on topics covered in the videos. I would estimate that in Fall 2015, until the first exam, a number of students did not pay sufficient attention to the videos. However, after the first exam, students began watching the videos more diligently.

One reason we flipped the course was to restructure class time so that students could spend more time in mentored environments working in small groups solving problems. As it turned out, though, students requested more lecture than the once-weekly format. Students struggled to grasp some concepts from the videos. While students can review these topics multiple times, I believe they sometimes needed an alternative explanation. In a lecture, when students ask questions, I try to respond with a different perspective or explanation. With the flipped model in Fall 2015, students had only one class meeting each week to ask me questions about the homework. The second time we ran the course, in Fall 2016, we had two lectures each week, and I think students appreciated the additional lecture time.

I really enjoy teaching this course. It’s a lot of fun and a great privilege. Many non-majors enroll, and humanities undergrads have shared that this was the first math course they enjoyed and they were impressed with the applicability and universality of statistics. The class typically enrolls about 100 students.  Even with this large number I am able to learn most of their names by the end of the semester when we met three times per week. I did feel, though, that I was not able to get to know students as well when we met once per week. More important, I think the once-weekly lecture deterred students from coming to see me during office hours: I noticed a sharp decrease in the number of students who consulted me during office hours in Fall 2015. In Fall 2016, under the twice-weekly lecture model, I had better office hour attendance and was better able to get to know students.

While we were happy with the increase in the number of lectures, I think it’s important that we not decrease the number of small group meetings. The worksheet activities were important for their learning. Students were not always as enthusiastic about the small group problem solving, but they adjusted to the format and things improved as the semester moved forward. Furthermore, we still found it better than a TA solving demo problems for the class, especially in terms of class engagement and in terms of fostering independent problem-solving.

We used two types of problems in the course. The first required more synthesis-based understanding of previous topics. We began to develop more basic, conceptual worksheets once we saw students were not always able to keep up with the videos.

We did not give students the solutions to the worksheets. We worried that if we provided full solutions, they might be less motivated to work through challenging problems and/or skip discussion section altogether, and participation in section was important. Students did get feedback from the TA when they presented their solutions in class, and we did provide solutions to most assigned homework problems.

Overall, we did not see a dramatic change in student learning. We did not conduct a controlled study of learning gains, but exam scores were not much different from year-to-year. Course evaluations for the one-lecture-per-week format were slightly lower. (Again, the main complaint was that students wanted more time with faculty member in lecture.) Students were happier with the two-lecture-per-week format we implemented in Fall 2016. Therefore, we plan to stick with this format, meeting four times per week so students attend two lectures and two small-group sessions per week. We have also been more explicit about the role of each component of the course – videos, lecture, clicker quizzes, small group meetings – and what students are responsible for completing and when.

Most of all, we were very lucky to experiment with this approach with many terrific TAs—we owe them a real debt of gratitude for their assistance. We gratefully acknowledge support from the Office of the Provost and President for a PILOT grant that assisted us in implementing the flipped course.


Avanti Athreya is an Assistant Research Professor in Applied Mathematics and Statistics (AMS) at Johns Hopkins University. Prior to flipping the statistics course, she and Professors Naiman, Fishkind, Torcaso, and Jedynak (all AMS faculty) implemented a case-study based approach to introductory statistics as a part of the JHU Gateway Sciences Initiative. Her research interests are in probability and statistical inference on random graphs.

Dan Naiman has been on the faculty in Applied Mathematics and Statistics since 1982. Upon arrival at JHU, he taught Statistical Analysis I for 3 consecutive years, and has continued to teach the course occasionally, as well as a host of other statistics courses at all levels, since then.

Image Source: CC Macie Hall 2013