To strengthen the connection between the statistical community and National Science Foundation, we continue the series introduced in the May 2023 issue of Amstat News that poses questions to NSF program officers and awardees. This month, awardee Po-Ling Loh tells us how she earned her CAREER award. If you have questions or comments for the program officers, send them to ASA Director of Science Policy Steve Pierson at  pierson@amstat.org.

Po-Ling Loh is a statistics professor at the University of Cambridge Department of Pure Mathematics and Mathematical Statistics. She earned her PhD in statistics from the University of California, Berkeley in 2014 and her research interests include high-dimensional statistics, robustness, and differential privacy. She is a recipient of an NSF CAREER Award, an Army Research Office Young Investigator Award, an Institute of Mathematical Statistics Tweedie New Researcher Award, a Bernoulli Society New Researcher Award, and a Hertz Fellowship.

What is your amount of experience with the NSF proposal process?

I was at the University of Wisconsin-Madison when I received the CAREER award in 2018. I submitted several collaborative NSF proposals in the past and sat on a handful of NSF selection panels from 2014–2019.

How will the funding be used?

The goal of the project, “Something Old, Something New: Robust Statistics in the 21st Century,” was to explore new theoretical directions in robust statistics. This includes the effects of high dimensions, adversarial contamination, and non-IID data, all of which are becoming more commonplace in modern data analysis but were largely absent from classical studies in the field. The funding amount was $400,000, which is standard for a CAREER grant, providing support for one PhD student per year.

Summarize the goal of what the proposal will accomplish.

Specific questions to be explored regarding high-dimensional data included the following:

• How do existing notions of robustness apply to high-dimensional settings?
• How should high-dimensional estimation procedures be modified to protect against deviations from distributional assumptions?
• How might one quantify the relative robustness of various proposals?

As many natural robust estimators involve optimizing nonconvex objective functions, the project endeavored to make novel theoretical advances in both statistics and optimization theory. The project also aimed to address some long-standing open questions in robust statistics involving optimization of low-dimensional nonconvex objective functions. Finally, the proposed work involved studying consequences of the new theory in machine learning applications such as medical imaging.

What impact has this funding had on your career and work?

Preparing the CAREER proposal helped me develop a cohesive five-year research plan, which was a useful exercise as an early-career faculty, even though I didn’t end up pursuing the exact projects outlined in the proposal—one’s research trajectory is not always predictable! The funding was useful because it gave me a stable funding source for five years, which is essentially the life cycle of a statistics PhD student in most American universities. Actually, the student who ended up being funded on this project was in the computer science PhD program, so the support was even more critical, since the funding model is different there.

What advice do you have for others applying for NSF funding and the specific program to which you applied?

Compared to other engineering disciplines, NSF CAREER awards in statistics are rarer; hence, they are more prestigious. However, this may cause potential applicants to be discouraged from even applying. I don’t think it hurts to apply for a CAREER award, since preparing a five-year plan can be helpful in setting a more coherent research agenda for the immediate future—it also makes things easier when a prospective PhD student comes looking for potential projects! Furthermore, the feedback from the reviews is more extensive than from some other funding agencies and can be quite useful. It’s also good to remember that applicants can make up to three attempts and previously declined CAREER proposals will not adversely affect future chances.

That said, the investigator’s track record is fairly important—as I witnessed firsthand from sitting on a CAREER panel—so the likelihood of receiving an award may be somewhat higher if one waits to apply several years into the tenure track.

I would also like to mention that many universities have centers or dedicated staff members who can help brainstorm ideas for writing the “Broader Impacts” section of the proposal. Often, this is the part in which investigators draw a blank on writing something innovative. I attended a workshop at UW-Madison specifically focused on helping write about broader impacts for NSF CAREER proposals and was inspired to be more ambitious and creative with this portion of my proposal. We were also told that if we were having trouble coming up with ways to turn scientific work into broader impacts, we were welcome to make an appointment with one of the staff members to discuss initiatives going on at the university and how to plug in.