University statistics courses – I’m not just an employee, I’m a customer!
(I bet none of you remember that Hair Club for Men commercial and don’t get that reference. That’s okay, I amused myself.)
In my spare time, of which I have none, I teach a statistics course once or twice a year. Maybe it is because I have made my living as a statistical consultant and statistician for the past thirty years. Maybe it is because I get tired of interviewing people with degrees who can’t DO anything, but I have a different view of what should be taught in statistics courses than some of my colleagues in the teaching profession.
We are all very polite to one another, but I suspect they are thinking,
“She doesn’t teach her students to understand anything. It is all done by a computer and it is all a black box to them.”
I, on the other hand, am thinking,
“None of what they focus on will the students ever use after this course. I’ve been using Analysis of Variance and regression for thirty years and the number of times I have needed to calculate an F-statistic using a calculator is exactly zero.”
I agree with Wolfram’s idea of Computer-based Math, that math does not equal computation.
It is not completely true that I don’t teach my students to understand the equations that are underlying the results SAS or SPSS or R or whatever application produces. I do think it is important that they understand an F-value is the ratio of two estimates of variance. In fact, I wrote a post on exactly that.
I am unconvinced, though, that hours of tedious computations drive that fact home. Nor am I convinced, after many years of teaching, that having students do those tedious computations for hours makes ANOVA or factor analysis or whatever any less of a black box. They simply memorize a formula, plug numbers into it and forget it all the day after the final exam.
Some of my colleagues may argue that students will need the knowledge of those formulae for some other exam they may take – an AP statistics exam for high school students, or their written exams for their Ph.D. I have never had a student fail their masters or doctoral exam, so I’m not too worried. Since my students these days are past the AP level, I cannot speak from personal experience.
I do hear from math teachers I respect a lot, though, that they do, in fact, require a lot of computation simply so students can pass standardized tests. I find that really unfortunate.
It’s ironic, really. I did my dissertation on standardized testing in the U.S. and Mexico and at the time I thought standardized tests didn’t warrant all of the strenuous objections to their use. Now, I am concerned that testing has perverted our educational system. If you are teaching something because it is on a test, even though you haven’t used it in your entire career in that field … well, it just seems wrong to me.
I completely agree with Wolfram that if you want to compute derivatives or logarithms or invert matrices or find the anti-log without using a computer, that is perfectly fine, just like learning Latin is perfectly fine. Like Latin, learning any of those things certainly has some value. I would find it hard to argue, though, that they are necessary.
Note: There is a huge difference between computation and understanding. I did not say you didn’t need to understand derivatives or logarithms. In my experience (not a random sample), really good statisticians aren’t particularly outstanding at arithmetic. Some are. Some aren’t. There appears to be little correlation. Perhaps my colleagues who require loads of homework problems with calculations done by hand would disagree.