You can take the long, hard, and expensive way, or use statistics. That may seem contrary to most students struggling through their first stats class, but statisticians actually work within many disciplines, applying their predictions in order to save time and money.
Shane Reese, of the Statistics Department, is well-versed in statistical models that do just that. He was recently awarded a grant from the National Science Foundation, which will fund his development of statistical models that will improve climate predictions through a better understanding of the magnetosphere, the upper layer of the atmosphere.
Reese’s statistical models will replace the long and costly computations that are normally done by supercomputers. His new time- and cost-effective models can lead to better predictions of the weather. Eventually, models of the earth, ocean, and air can be combined to obtain improved models of climate.
Though Reese hopes to eventually impact these future climate models in the long run, he is focused on getting the preliminary statistical models for the magnetosphere in place.
“This goal [of better climate models] is a long way down the road,” said Reese, “but in the short-term outlook, our goal is to predict well, to validate, and to verify.”
Reese is meeting two challenges in creating reliable predictions. First, it is computationally intensive work on the statistical end. The second challenge in the project is the fair amount of learning required to understand the physics involved in the problem.
“The challenge is to understand enough of physics to be conversant,” said Reese. “Statisticians get to play in everyone’s backyard.”
Once an understanding is reached, Reese can develop effective models in order to guide actual experimentation.
“It’s important to figure out cleanly what we know and what we don’t know,” he said. “Statisticians are good at quantifying this statistical uncertainty, and then using it to better inform the process of collecting data.”
Reese has previously shown his capability in reducing cost and time by informing, within several other disciplines. For example, he co-authored a study of statistical models that allow the manufacturing industry to evaluate the effectiveness of tests. Using the models, companies can make more informed decisions between tests which require complete product disassembly or destruction and other non-destructive testing methods. This study, published in the Journal of Quality Technology, provided the industry with vital and cost-effective information that was previously absent. Hopefully, Reese’s current research into the magnetosphere will yield similar positive outcomes.
At BYU, Reese is also currently working to prepare more statisticians to contribute. Thanks to a grant from the Dean’s Office, students can now pursue mentored research in the Statistics Undergraduate Research Computing Laboratory. Reese spoke highly of the advantages of the new lab.
“This is as good as it gets for a statistician—just a dedicated space to do research.”