Statistics professor Len Stefanski created computer programs in December that allow him to hide images and messages in data sets, and began to formally promote his programs in April.

Stefanski’s programs served to make the end results of solving a correct multiple regression analysis on data more interesting for students, he said.

According to Stefanski, what he does is after creating a data set, he gives his students a homework problem and then asks them to find the correct multiple regression analysis on the data.

“If they are successful and they do a final step of checking their results using a residual plot, then they’ll see the message or the image,” he said.

He said the program is still fairly new and only a few of his students have tried it, but he has gotten a positive response from the ones that have.

According to Stefanski, multiple linear regression analysis is a concept that most students learn in beginning statistics courses, a method that is used to show how one variable depends on another.

“Usually, that last step [in the analysis] is a meaningless pattern,” Stefanski said. “So, I found a way of doctoring up the data ahead of time, so you get a little surprise at the end.”

Jeff Thompson, an assistant statistics professor, used Stefanski’s programs for a graduate class during the spring semester and for one of his assignments, Stefanski scanned in a picture of Thompson, so that if Thompson’s students did the correct plot for the data set, a picture of Thompson would pop up.

“It just makes a funny sort of punch line at the end of an assignment,” Thompson said.

He said anything that can lighten up a lecture or that his students may find funny is worth a try.

Stefanski began working on this project in October 2006 and in December, after having surgery, which he said caused him to be bed-ridden for a couple of weeks, he worked on writing the computer programming and finished his research and writing of a journal article. The American Statistician accepted his article in January for its May issue and the editor of the publication called Stefanski and asked him for permission to promote his work, which Stefanski agreed to help with.

Thompson said the method “relies on some heavy computational calculations” and therefore has to be done on the computer. Since his ST 512 class had a lab component, he said it was perfect.

“[Finding an image or message in data is] just something that’s unexpected at the end of a problem, a bit of a surprise, if you will,” he said.

On Stefanski’s Web page, which he continuously updates, instructors can click and download data sets to use in their classes. He said one of the images he used was of Homer Simpson and that he had to get permission form the University’s legal department to make sure he was allowed to use it because of copyright infringement laws.

Depending on what kind of market is available for it, Stefanski said he wasn’t sure whether or not he would patent his project.

“For right now, I’d be happy if other instructors around the country [used it],” he said.

A professor from Germany, Stefanski said, e-mailed him saying he was planning to use it in one of his courses, and another professor from Harvard said he implemented it in one of his classes.

Stefanski said a programmer at Statistical Analysis Software wrote a version of his program where the SAS programmer added color to some of Stefanski’s plots, which were all black and white.

“I haven’t looked too much into it yet,” Stefanski said.