A colleague, Scott White, pointed me to this article in the Guardian: What do you get if you don’t teach stats properly? Farage and Trump.
The author, Robert de Vries, is a lecturer in Quantitative Sociology.
He argues that we are not teaching our students to think carefully about the meaning of numbers:
But we’ve been trying to [teach the fundamentals of statistical analysis] for decades and it just hasn’t worked. Instead we have run course after course that students hate. We’ve turned out generations of graduates who can remember sitting in labs pressing buttons in statistical software programmes like SPSS, but never really learned how to connect statistics to important issues in the real world.
I believe that our fundamental problem is that we do not expect our students to really understand what we teach. We leave them with a makes sense epistemology; they see that we don’t expect them to understand, so they must accept what we say. When they think about statistics they have a sense of helplessness and passivity. They think “I guess so” or “I’ll take your word for it”. They must think that way; that is how we taught them.
Statisticians have also been worrying about these ways of teaching. George Cobb argues that there is no practical way to teach the curriculum as currently taught, and expect the students to understand it at anything but a very superficial level (G. W. Cobb 2007; G. Cobb 2015). We have to change our focus from the older mathematical tests, such as t-tests and correlation, to tests based on resampling, such as permutation tests. These tests use ideas that are much simpler to understand, so the student can reasonably hope to get a deep understanding of fundamental ideas such as sampling distributions and statistical inference.