During a lecture, he pulled out a bag of cookies and suddenly the students paid attention, he said. Then he gave a cookie to each student and asked him or her to count the number of chips (which requires carefully eating the cookie, since not all the chips can be seen from the outside). Lee graphed the data, showing a Poisson distribution, which is close to a classic bell curve in this case.
"The number of chips match what we expect statistically, but not what we know intuitively," Lee explained. "The students don't expect individual cookies to be that different, but they are. This example shows that variability is all around us, even in something like mass-produced cookies."
For instance, when it came time to study two-sample t-tests, Lee brought in a second brand of cookies. The students got excited, he said, as they tried to figure out how to compare the difference in brands in light of all the variability between individual cookies. And, of course, they got to eat more treats.
Lee has even found chocolate chip cookies helpful for teaching graduate students about Bayesian analysis, a method for dealing with uncertainty that requires starting with an educated guess, or "prior."
"It's interesting because the students' initial guesses on the number of chips are usually very wrong," he said. "They find out what happens when their guesses are fairly far off, and how much they can update their prior by collecting real data."
Lee now has developed protocols that he plans to use in his statistics courses for the foreseeable future.
"If I hook students with the cookies, I find they keep coming to lectures for the rest of the course," he noted. "It has worked so well, now I use the cookies as much as I can."
-Chocolate chip cookies help make statistics lessons relevant and palatable