2014 ESRA Winner Talks Statistics, Journalism, and P-Values

Morteza Marzjarani

When did you become interested in statistics, and who inspired you?

When I was a college freshman at the University of South Florida, I thought I wanted to major in chemical engineering, so I took a job in a polymers lab in the chemical engineering department. But it turned out that bench work and I didn’t get along. I’m just no good with my hands at all. So, the lab director had me work on a data-analysis program in the FORTRAN language instead. When he explained the least-squares-error minimization concept behind the program, I thought, “Aha! This is so much better than mixing chemicals!” So I switched to industrial engineering. I took classes on applied probability and statistics—queuing theory, quality control, stochastic processes, and design of experiments. I loved it so much that decided to get a PhD in either operations research or statistics. I asked an engineering grad student for his opinion. He said, “Don’t do statistics. Everything that can be done in statistics was already done by the 1950s. It’s a dead field.” This was 1995. I decided to bet on him being wrong. Thank goodness I did. I started the doctoral program in statistics at Stanford University in 1996, right at one of the most exciting times to be in Silicon Valley and study statistics. Best decision of my life.

Regina Nuzzo is the 2014 winner of the Excellence in Statistical Reporting Award and professor of statistics at Gallaudet University in Washington, DC. She earned her PhD in statistics from Stanford University in 2003.

Statistics and science journalism don’t seem like they have anything in common. Was it hard to transition from doing statistics to writing about science?

Not as much as I expected, actually. Science journalists and applied statisticians are kindred spirits. Both require some of the same skills. You need to be able to synthesize and analyze a large amount of information, and boil it down to its essence. You need to communicate complicated ideas in a way that’s easy for nonspecialists to understand. And you’re trained to have a healthy sense of skepticism. Journalists have a saying about the importance of gathering evidence for a story and not trusting anecdotes: “If your mother says she loves you, check it out.” To me, this sounds like something a statistician would say: “What’s the evidence that my mother loves me?”

I see you work at Gallaudet University in Washington, DC, the nation’s only liberal arts university for the deaf and hard-of-hearing. What is that like?

Gallaudet is a fascinating place to work. Everything on campus is done in either American Sign Language (ASL) or written English—no spoken English. I teach statistics to undergraduate and graduate students in sign language. It’s fun and also very challenging. ASL doesn’t have signs for all the statistical concepts, so you need to be creative in how you teach things and really make everything concrete and visual. How do you make the concept of standard deviation, or eigenvalue decomposition, or moment-generating function succinctly visual? I get four dimensions to do that—three spatial, plus time—so it’s kind of fun to play with.

I didn’t know sign language when I started teaching at Gallaudet. I was born with a severe hearing loss, myself, but I used hearing aids and lip reading growing up. Many students with hearing loss use sign language interpreters in college and grad school, but when I was at Stanford, I used the university-provided real-time captioning for the biostatistics seminars, kind of like you see on TV. Except, here, the captionists had to figure out how to spell words like “heteroskedasticity.”

Is it hard to do your writing and also be a university faculty member?

It’s not easy, to be honest. University systems are traditionally centered on a teaching-and-research rewards model, and the sort of statistics communication and science communication that I and others do doesn’t really have a formally recognized place in that. I have colleagues in other fields who are in the same boat—they have PhDs and teach, and they also write award-winning books, essays, and articles and do public speaking. But sometimes it’s hard for universities to know how to weigh the value of this compared to publishing a technical paper in a peer-reviewed journal. I think they’re both valuable in different ways. What’s really intriguing is the UK’s model of “professorships in the public engagement of science,” or the “public understanding of science,” where a faculty member’s formal duties involve these sorts of science communication activities. It would be interesting to see how this model would apply to the United States. Wouldn’t it be exciting to have a U.S. professor for public engagement in statistics and data?

What do you think about ESRA?

I’m so excited that this award exists! There’s so much room for data and statistics in modern journalism—data visualization, investigative reporting projects that come from original data, plus covering all the new developments in statistics. This award is a terrific idea and a terrific honor. I can’t wait to see the 2015 award winner.

What have you done since you won the award?

I’ve written a few more statistics-themed articles, including a news brief for Scientific American on nonrepresentative political election polling, a feature for New Scientist on the state of the Bayesian-frequentist debate, a technical article for a medical journal on the inverse fallacy and misinterpretations of the p-value, and a news story for Scientific American on a psychology journal’s decision to ban null hypothesis significance testing. I’ve also given a few talks about my Nature feature and statistics in the media to different audiences, which has been very enjoyable.

Your article about p-values won the 2014 ESRA award. In a few sentences, could you please explain your view of the p-value and what motivated you to write the article about the p-value?

The idea for the Nature feature grew out of some reporting I had done for a shorter Nature news piece. I realized that between the great work that was being done on p-hacking, the common misinterpretations of p-values, and the reproducibility crisis in science, there was a very interesting story here. I worked with my fabulous editors at Nature to really find the angle for the article.

My own view of the p-value? I say it’s like a screwdriver—designed for a particular purpose, and it’s very good at doing that, driving screws, but the problem comes when we use it to bang nails and then complain that it’s not a good hammer. P-values are misunderstood.

But if this article makes nonstatisticians a little more curious about what’s going on underneath the hood of p-values, then that can only be a good thing. I’ve had so many people contact me to say the article has sparked their interest in the whole issue, which is pretty exciting.