In Part I of Becoming a Great Teacher I discussed how the single most important attribute of great teachers is that they are dynamic presenters. What do I mean by dynamic presenters? Among the descriptions students provided, two are: doing something other than lecturing and keeping students entertained.
In this entry, I suggest one method of breaking the monotony of the classroom I use for a statistics class I teach. The goal of the class is to teach students how to make sense out of large data sets. On the second day of class, to give them a fun data set to begin with, one where they generate the data themselves and therefore understand the data intimately, we leave the classroom and play sports.
First, we go to an open field to hit softballs (picture on left taken today). Each student is expected to swing at each pitch until they hit three balls. Letting the balls roll as far as they can, we then measure the hitting distance. During the next class we look at the data and observe that some people hit very far and some people not far at all. After asking students why some people hit further than others, they quickly volunteer variables like gender and experience playing baseball or softball. At that point, we collect the variable information for each player and I teach them how to run a regression with hitting distance as the dependent variable and variables like gender and experience as the independent variables. The data typically work great, with predictable coefficients and statistical significance.
When teaching how to use test-statistics, I ask the students for the last digit of their cell phone number, and then use that digit as an explanatory variable. It is absurd, of course, but they are surprised to see that the regression will not estimate a coefficient value of zero for this variable. Then they are introduced to the concept of a test-statistic, and of course, the variable is always insignificant. I have been doing this for six years, and have collected over 400 observations. Readers are free to download these data by going here and clicking on Data Available For Download in the left menu.
An alternative to softball is basketball, which I have substituted for softball if it is raining, and this year did it concurrent with softball (picture on left also taken today).
Each student is asked to attempt five free throw shots and five three point shots, and to record the number of shots made each time. We then utilize the number or percent of shots made as the dependent variable, often as a function of gender, experience, and a dummy variable for the three point shot. As you would expect, the coefficient for the three point shot is usually negative and significant.
A Forecasting Contest: At the end of the semester we hold a forecasting contest. Using data collected at the beginning of the semester, students form teams and develop a regression model forecasting softball hitting distances or basketball shooting percentages. The class then returns to the field/court, and are required to bring a guest. Each team predicts the performance of each guest in hitting softballs or shooting basketballs, then the guest is asked to hit/shoot. The teams with the lowest forecast errors receive a higher grade. This activity provides a concrete demonstration of out-of-sample forecasting. Also, because more parsimonious models always seem to win, it illustrates the advantages of simple but smart models.
Readers are also free to download the basketball data I have collected by going here and clicking on Data Available For Download in the left menu.
This activity is not only fun, but I believe helps students better learn regression. Whenever I introduce a new concept in the class I always use the softball/basketball activity as an example, and they seem to catch on quickly. Moreover, students love a break from the classroom.
I've always said I don't know if I know how to teach, but I do know how to get high evaluations, and this activity is guaranteed to boost your evaluations!
