Empirical studies on teaching statistics, or Why can’t we all be like Eric Mazur?

Kai Ruggeri writes,

I am doing my PhD research on teaching strategies for statistics and have been reading through your book. Regarding your first day/week activities, I was just wondering if there were any empirical studies done to demonstrate the productivity and/or benefit of these exercises. I have had a great deal of trouble in searching for certain empirical data.

My reply:

There’s been a lot of research in effectiveness of active learning (see, for example, McKeachie’s Teaching Tips or various other books for teachers). In physics, Eric Mazur has developed some tools called Peer Instruction, which look pretty cool–there is some similarities to the methods in my book with Deb Nolan, but much more developed. And he and others have performed some studies on their effectiveness. So this might be a good place to start. For example, Mazur writes,

Demonstrations are universally agreed to be the fun part of physics classes. But do students actually learn much from them? The answer from education research seems to be no!

We are searching for ways to make demonstrations more effective by asking students to predict outcomes before seeing the demonstration, thus forcing them to think about the physics and enhancing their interest in the demonstrations. Careful research will tell if students do learn more from demos presented this way.

I’m planning in a few years to do more development on our active learning project as part of revamping the introductory statistics course, and when doing this, I’d like to perform some evaluations with pre-tests, post-tests, attitude surveys, and so forth. I’d been planning to do this already, but other projects got in the way, and I also felt dissatisfied enough with the standard course to decide that I’d like to alter the material covered in the class as well as the way of teaching it.

3 thoughts on “Empirical studies on teaching statistics, or Why can’t we all be like Eric Mazur?

  1. I have found the case method to be well suited to teaching statistics since it is in many ways an art. By having students discuss different ways of doing things, it helps them gain insight into how to think statistically and most importantly to understand that there are many ways to model a situation. I guess I'd classify that under active learning too.

  2. I have each student from the begiining of the semester…working on a different data set. Then as we progress through the course…plots, summary stats… confidence intervals…they submit an example of their work to a public online portfolio…
    Then they comment on each other's efforts…they are allowed to edit as they learn more…this is simulating peer reveiw and workplace accountability. Then they finally write a term paper summarizing everything they did.
    This paper is the main deliverable…the most points.

  3. professor gelman,

    i was curious how you would alter the material covered in a typical intro stats course. what topics would you discard? what would you keep? what needs to be added?

    aleks, what do you think?

    i think it is analogous to the idea in this other post.
    http://www.stat.columbia.edu/~cook/movabletype/ar
    where practically all introductory stats books and courses cover the same material. whether it is due to orthodoxy or some notion that some topics are intrinsic and must be taught, who knows?

    jimmy

Comments are closed.