or years, I have been telling my journalism students at New York University that I had a little secret when it came to deconstructing scientific papers I wrote about: I kept a statistician in my back pocket. I told them they should do the same.
My back pocket is a bit empty today. Last week, Howard Barkan, the biostatistician I kept close, died after suffering a massive heart attack while giving a lecture at the University of California, San Francisco. He was 70.
And in reflecting on his legacy, I’ve been reminded of the valuable, and underappreciated, role of the numbers-minded sort in keeping science honest. As data sets grow ever larger and the statistics required to interpret them ever more complex, statisticians — working with scientists, granting agencies, and journalists — are undergirding our scientific findings with a sound mathematical foundation.
It’s only fitting that I should write an obituary of sorts for Barkan, since we first crossed paths a little more than a decade ago while I was writing an obituary for the Lancet about someone else — Henrik Blum, a longtime professor at the University of California, Berkeley, who died in 2006.
I frequently end interviews by asking a question like, “Is there anything else I should have asked?” but Barkan flipped the script. Tell me about your work, he asked, in the sort of genuine way he showed interest in people, and I did. He wondered whether there was some way we could work together.
It turned out that there was. I had recently begun writing a column called “Statistically Speaking” for a group of publications for health care professionals, and while I was getting great reader response, my readers also pointed out a few errors along the way. They were small — a few miscalculations or assumptions, that sort of thing, quickly corrected — but my editors and I wanted to make sure the copy was as clean as possible moving forward. Was there anyone I could consult for feedback and advice?
And so Barkan became my statistical backstop, teaching me invaluable lessons with each phone call and email. That’s who he was, a generous and patient teacher who loved imparting wisdom and helping others improve their own work. “Those who needed help setting up a study or finding the correct statistical analyses could count on him to be ready with an answer and a smile,” said Jerome Minkoff, who worked with Barkan in the mid-1990s on studies of the effects of abuse on women’s likelihood of having mammograms and other preventive care.
Barkan “always advised the students to jump right in and ‘play’ with the data and the analysis program to understand how it worked and where they had questions,” said Dr. Karen Sokal-Gutierrez, whose Berkeley students worked with him.
Barkan was like the Zelig of statistics and epidemiology. (Another colleague, Melissa Farley, describes him as a Renaissance man; he mounted photography exhibits around the Bay Area and frequently played guitar in shows around the Chicago area when he lived there.) The last time I visited Barkan in Berkeley in March, we had a typically wide-ranging conversation, alighting on the lessons of the Tuskegee study and the FDA drug approvals process. He reminded me that he had been involved in studies that had sounded the alarm about a substance called hetastarch, used during heart surgery since it was first approved in the 1960s. Half a decade after Barkan and his colleagues began publishing their work, hetastarch was at the center of one of the largest scientific fraud scandals of recent years, featuring 94 retractions and earning a “black box” warning from the FDA in 2013.
The fact that in that case more people should have listened to the biostatistician and his colleagues is no surprise to any of us who’ve followed science closely.
“Biostatistics is blooming,” Stanford Medicine magazine proclaimed in 2012, and it was right. Statisticians’ work becomes only more important in the era of Big Data. And yet statisticians “are like a cog in a bigger system,” as Ben Goldacre put it last year. “I would hope that statistics is becoming an integral part of science,” Goldacre told Statistics Views. “When you look at how statistical tools are routinely misused, it would be great if it was more thoroughly embedded.”
That was the plan with a new project Barkan and I started discussing in March. We envisioned a blog and a numerical database, bundled together, the statistician embedded. Sadly, he will no longer be, but we will dedicate the project to him, in honor of the biostatisticians in all our back pockets.