I recently got back from the Joint Statistical Meetings in Washington, D.C. where I talked about making audiences concrete and motivating authentic arguments for statistics students (and spread the word about MathStatBites of course). This is a big conference where statisticians from all over the world get together to talk shop, and it was back in person after a few years of going virtual.
Regina Nuzzo, a science journalist and professor at Gallaudet University, was the discussant for the Teaching Statistical Communication session that I was presenting in. As part of her remarks, she got meta and noted the need for us to practice what we were preaching in the session. How might we share our work not only with one another at this conference but to a broader audience? One approach is to translate past JSM abstracts for the public which she had experience doing from past meetings; that sounds in the spirit of “bites” to me. This call to action got me thinking, so here is my translation attempts for two talks that I attended.
Robust Inference for Matching Under Rolling Enrollment – Amanda Glazer and Sam Pimental
When we want to understand the effect of some treatment, we need to compare people’s response for those who had the treatment and those who did not (let’s call this group, control). But what if a person’s response is not really due to the treatment but rather due to some other quality about them, like their family history or their sleeping habits? If people were randomly assigned to treatment, then these kinds of differences should wash out on average, but if we are just observing people who fall into treatment and control categories, we might think to compare people who were alike in all ways except for whether or not they were in the treatment group. Then we can hopefully isolate the effect of the treatment. However, this “matching” process can be challenging when people are coming into the study on a rolling basis rather than all together at the beginning. This talk presents a case study using data on major league baseball players where the “treatment” is an injury and the “response” is the batting performance, and shows a way for matching to occur across time by matching players at one time point with other players at a potentially different time point where they are most similar. Batter up!
Why does linear regression “work”? What makes the normal distribution so special anyway? Answering these types of questions, ones that underlie what you might learn in an introductory statistics course, often requires two calculus-heavy courses. Even though introductory statistics courses have been changing to de-emphasize calculation and re-focus energy on understanding the concepts, upper-level courses are slower to change. Might there be a way to move away from all of those nasty derivations and focus more on the problem-solving aspects of the theory? But if we are changing the focus of the course, we must also change how we will assess whether students are learning. This talk shares activities and assignments that double down on concepts over computation.