Florentina Bunea wants to understand uncertainty.This philosophical curiosity is what brought her to her current field of work in statistics, unlike her more prosaic peers. “You typically arrive at statistics to solve big data sets and solve big problems,” says Bunea. “But my initial motivation was: How do we quantify uncertainty? I found it fascinating.”
Her fascination with pinning down the unknown has led to her research in theoretical statistics, where she works on foundational problems that can be used down the road in practical applications. Her primary focus deals with teasing out low-dimensional structures in high-dimensional systems so that statistical analyses can be performed on those systems. When researchers have a small sample size, but each data point has a wealth of different variables, drawing statistical conclusions can be almost impossible. Bunea analyses models that can tease out information from these ‘impossible’ data sets.
A good example, says Bunea, is in neuroscience, where researchers want to understand the connections between regions of the brain. “The brain is a complex structure,” says Bunea, “but in some cases, say if you’re studying a neurological disease, perhaps it’s enough to study a sub-network. Can we draw a conclusion about the whole system using this smaller network?” Another real-world example is determining what genes cause a certain type of cancer. “The challenge is,” says Bunea,” you’re looking at fifty thousand, sixty thousand genes, to find the ten that are responsible. You have to prune it down.” This statistical conundrum pops up a great deal in biological and life sciences—and Bunea’s work helps to solve it.
Her work in the field was put to the test when she was asked by researchers at Brown University to help statistically analyze the quality of life of patients living with HIV and Hepatitis C. The researchers had a small sample size, but each person in the study provided a great deal of data—demographic information, clinical information, as well as test scores from a battery of neurological tests, and neurological images from fMRIs. Of all these variables, the neurologists wanted to know which ones were most associated with lower scores on the neurological tests, and which areas of the brain should they focus on when studying these patients. “They already had an idea of which areas were important, but they needed statistical analysis to confirm it,” says Bunea. “Our work help confirm what they had observed in the lab.”
Bunea chose to come to Cornell thanks to the opportunity to contribute to the relatively new Department of Statistical Science. “There has been a long tradition of statistics at Cornell, with prestigious graduates, and Nobel prizes,” she says, “but the statisticians were spread throughout the campus, and only recently did they come together. That’s what attracted me, the creation of a new department.”
When she’s not running statistical models, Bunea is running after her toddler, a son she has with Marten Wegkamp, also a professor in the department. “I used to be a very serious movie-goer,” Bunea says, adding that she’s also fond of contemporary interior design—a statement backed up by an enormous clock that hangs on her office wall, splashed with vibrant, modern patterns. “I bought it for my son’s room and realized it’s not exactly an infant-type object—maybe I’ll give it to him later,” she says with a laugh.