Learning New Statistical Tricks

Cran, sass, win-bugs, jags, arm, and arr: these nonsense words might blow over your head unnoticed. But if you’re a statistician, you’ll recognize these sounds as the names of computer programs used to quantify risk. At the 2012 BYU Summer Institute of Applied Statistics, held from June 13 to 15, these words buzzed as attendees learned how to make these programs work for them.

The Summer Institute, now in its 37th year, offers participants a chance to learn new statistics methods. This year Gilbert Fellingham, the newly appointed associate chair of the BYU statistics department, taught participants how to use applied Bayesian statistics, a form of statistics that is rapidly gaining followers in the discipline.

“The purpose of this seminar,” Fellingham said, “is to get people to understand the Bayesian paradigm and to be able to use currently available software to do analyses in [it]. This is just an effort to bring people up to speed in a discipline that they might not have had any formal training in.”

Though Bayesian methods have been around for a long time, most statisticians have been trained in frequentist methods, a different approach to statistics.

“With the advent of the computer back in the fifties and sixties,” Fellingham explained, “frequentist methods were easier to computerize. Frequentists got this huge advantage because you could solve a problem. Even if you were doing it wrong, at least you got an answer.”

Though Bayesian statistics has a wonderful theoretical structure, universities didn’t teach it as a practical method until recently. Now that computer technology has sufficiently advanced to handle Bayesian methodologies, universities and statisticians are working to catch up.

“Modern Bayesian computational statistics is less than 20 years old,” said Fellingham. “No one had Bayes stuff in their degree programs because you couldn’t solve answers until just recently. Now most research faculty are doing Bayesian things, but it hasn’t had a chance to disseminate to people who went through school ten years, fifteen years, or twenty years ago.”

Those who attended the Summer Institute walked away with new techniques they can apply in their fields. Mike Anderson, a faculty member in the biostatistics department at the University of Oklahoma, came to the institute to expand his abilities within Bayesian methods.

“I’m hoping to come away with really hands-on experiences,” he said. “Sometimes you go to places and they have these discussions that are all theoretical. Here I’m going to come away knowing how to do some of this. When I see a data set, my knee-jerk reaction won’t be to go to my frequentist methods, but [I’ll] be able to say, ‘Hey I know how to do this in Bayesian’ and be able to apply some of these things.”

By Katie Pitts Posted on