Nicholas Horton asks:
Crowdsource request to statisticians our there: to make room for modern topics in intro stats what topics might you drop or de-emphasize? I welcome your ideas
No Math: For 95%+ of the intro stats students who don’t go on to statistics PhD’s, math is a waste of time, like learning ancient Greek. Of course, math, like Greek, can be cool and fun and educational. And there was a time when the professoriate thought that you could not be an educated man without a knowledge of ancient Greek. But those days are gone.
No Terminology: For the 50%+ of intro stats students who won’t take more statistics classes, time spent on technical definitions of “bias” and “efficiency” is time wasted.
No Frequentist Nonsense: Just because we have spent the last 100 years inflicting a fundamentally flawed approach on our students does not mean that we have to keep on doing so. Hat-tip to Don Rubin.
No Tests: Null hypothesis significance testing (NHST) is an intellectual cul-de-sac. Don’t waste class time on the endless litany of specific tests. Amateurs test. Professionalizes summarize. Hat-tip to Allen Downey.
No Lectures: The typical intro class wastes a huge amount of student time with lectures. Stop talking. Your students are not paying attention. They are web surfing.
This is purposely exaggerated. By “No” in the above, I really mean “90% less than you currently do.”
- This assumes a fair amount of control over your own curriculum, which I am lucky enough to have. If Department X uses your intro course as a requirement, they might complain, might insist that you teach the frequentist interpretation and mechanics of a t-test. Solution?
- Tell them to jump in a lake. It is your course and you have a responsibility to teach it in a manner that is best for all your students. If they think that X is important, then they can teach it.
- Write up a ten page paper with all the material they think is so important. Tell your students to read it. Devote 15 minutes of class time to it. Ignore it for the rest of the semester.
I would push back on the substance of Nick’s question. Just what does he mean by “modern topics?” As long as he means the same material which I cover in my intro course, then we are in agreement. To the extent he means topics like deep learning, then I am suspicious of the possibility/usefulness of including them in the first semester of a data science/statistics class. I don’t think that “modern topics” like that can be meaningfully taught until the second semester, and I am not even sure about that.