“I was super elated,” says Brian Kwan, upon hearing he had been awarded a National Science Foundation Graduate Research Fellowship—so much so that he couldn’t sleep that night. Receiving the fellowship—one of 16 awarded to statistics students out of 2,000 recipients—was the culmination of years of anticipation, during which he resisted the temptation to rush an application while an undergraduate and, instead, deepened his knowledge of statistics by working on projects with faculty members at the University of California, Irvine, University of Pittsburgh, and his current home, University of California, San Diego.
Quoting the line from John Tukey that “The best thing about being a statistician is you get to play in everyone’s backyard,” Kwan says collaborating with different investigators gave him a chance to learn about the statistical literature in each scientific field and to reason which statistical methods could apply best to the statistical analysis needed. Working with Loki Natarajaran on research to develop novel statistical approaches to predicting future kidney function decline among type 2 diabetics deepened Kwan’s interest in prediction modeling and gave him the confidence, he says, to apply for the fellowship.
Maria Jahja heard from a friend she had been selected, and after checking the email, she immediately texted her parents. “It was all typos because I was so excited,” she says. “I know so many great students who applied in my field, I didn’t think I really had a chance.” As an undergraduate research assistant in the lab of Eric Laber at North Carolina State University, Jahja began building artificial intelligence agents for video games. “I would code fun games, then implement learning algorithms for sequential decision-making under uncertainty.” This led to her research proposal on using statistically rigorous uncertainty measures to inform decision-making, which she will pursue at Carnegie Mellon University.
“Computers are astonishingly efficient at solving formal problems,” she says, “but building an intelligent system for uncertain situations is much more difficult. Human intuition and reasoning are hard for a machine to reproduce, especially in complex environments where even a human expert might be unsure what the best choice is. It might even be that there is no ‘optimal’ choice, as it depends on individual preference. But if we could make some hybrid data-driven system capturing the strength of both—an expert and algorithm-driven decision-making system—I think that has immense value for society.”
Derek Hansen came to his research proposal, in part, through working at the Federal Reserve Board as a research assistant, where he used state-space techniques to estimate models in economic and financial applications and wrote lots of code in R and Julia to tackle the technical problems encountered (he will present the results of this research at JSM in a presentation titled “Randomized Missing Data Approach to Robust Filtering with Applications to Economics and Finance”). His NSF proposal will look at whether similar techniques can be used to improve model selection.
“To be honest, I wasn’t expecting at all to win the NSF fellowship,” says Hansen. “In fact, I found out because a fellow research assistant at the Federal Reserve saw my name on the website and congratulated me. I was absorbed in [a] Bayesian particle filtering project, so I hadn’t even checked my non-work email in a few days.” Hansen says the news was both a pleasant surprise and reassuring. “It’s a little scary working on a research proposal and sending it off to be evaluated—I also am relieved that experts in the field of statistics don’t think my ideas are completely crazy.” Hansen will begin a PhD program in statistics at the University of Michigan this fall.
The NSF fellowships provide financial support for three years across a five-year period.
Leave a Reply