On ‘Statistics’ Class and Math.

To Calculate or Not to Calculate?

I adjunct as a statistics professor from time to time. I’m in my fourth round, and my class has changed a lot over the iterations. The major issue is always how to balance the mathematical basis of statistics with its practical application.

Mathematically, introductory statistics is easy. You do some probabilities, you sum over some iterations, maybe divide them. Once in a while you do some factorials. This all builds up to linear regression. Easy, right?

The students in my winter quarter class couldn’t do fractions. Really. They were not able to tell if (1/4)+(1/4) was larger or smaller than (1/4)*(1/4). This was my first foray into teaching a night class, and the students were older, many more years removed from high school algebra than my usual daytime students. So I dropped the math. All of it.

We moved the class into the computer room and it became a course on data analysis. They still had to do means, medians, and modes; but instead of knowing you can’t get a mean for categorical data because the formulas didn’t work, they had to learn that categorical data lacks order, let alone even iterations, so one can only take a mode. They had to pick the right tests, run them, interpret results. It worked. They ran up results for data, week after week, and I think some of them may have been able to apply it to the jobs they were already working.

This begs the question: is data analysis math? I don’t think so. You can use math to understand statistics, but doing data analysis has a lot more to do with knowing the ins and outs of what you are using than how it is constructed. We all use packages, and we don’t read the source code for all of them. But we ask our students to focus on the mathematical wonders achieved in statistics, sometimes at the cost of actually getting them to use statistics to do something. Data analysis is a process using statistics, and it might be a better requirement in social science than statistics currently is.