Karen Kafadar

The key word is react. The statistics profession’s reaction to data science prompted my colleague to ask this question: “Why does our profession often find itself in reactive mode—reacting to others’ direction—rather than being proactive and setting the direction ourselves?”

It’s a good question. In some sense, it has been the basis of our profession for many years. During a special lecture at Indiana University in March 2008, Peter Hall said, “Statistics is ‘reactive’: It is very responsive to new problems that arise in chemistry, biology, physics, …”

We’ve been fortunate to have advances in our field driven by others who bring us interesting data that prompt the development of new methodology. We did not have to create those new data types—others did, and then brought them to our attention. A great example of this took place during ASA President Jonas Ellenberg’s invited address at JSM 1999 in Baltimore, when Eric Lander described the tsunami of genomic data desperate for statistical methods and research. (Others may have been introduced to these problems in other ways.) It was a call to action, to which many in our community responded, leading to meaningful advances both in statistical research and genomic science. It generated productive collaborations and energy in an interdisciplinary field that we now call bioinformatics, which “combines biology, computer science, information engineering, mathematics, and statistics to analyze and interpret biological data” (Wikipedia!).

Bioinformatics presents a positive situation in which we stepped up to the plate presented to us. But do we have to wait for scientists to drop the problems on our laps? Or can we take another approach, one in which we create the opportunities—before others do and develop new disciplines and institutes to solve them before statisticians even realize what’s happening?

On a tiny scale, statistics created its market in forensic science, but, again, it came about somewhat by serendipity. The National Academy of Sciences (NAS) Committee on Scientific Assessment of Bullet Lead Elemental Composition Comparison, formed in 2003, needed statisticians. Most committees in NAS’s Division on Earth and Life Studies never included a single statistician; this one couldn’t avoid it because two of the four aspects of the charge clearly related to statistics. Cliff Spiegelman of Texas A&M and I were invited to participate, Cliff for his years of expertise with statistics in chemistry and chemometrics and I because of my knowledge of chemistry (from college, National Institute of Standards and Technology, and HP).

Seeing how much impact statistics had on that study, the study director for the Committee on Identifying Needs of Forensic Science Community, Anne-Marie Mazza, invited Constantine Gatsonis of Brown as co-chair and me as a member. The final report, Strengthening Forensic Science in the United States: A Path Forward, emphasized the importance of statistics in forensic science at a time when many forensic practitioners may not have known just what statistics is.

Maybe we can call that one pseudo-proactive: We moved on it, but only after statisticians were called in and, SURPRISE!, we added more value than our “customers” (the forensic community) expected.

Data science is another example, and statisticians are responding in different ways. Some are waiting to see if “data science” goes the way of expert systems and artificial intelligence—now seen as largely oversold—and continue to conduct statistical research needed for sound inferences. Some academic departments have changed their names to “Statistics and Data Science” in hopes of directing the data science students and consumers in our direction. And others are ready to abandon traditional statistics altogether and jump on the data science bandwagon.

David Donoho wrote in the Journal of Computational and Graphical Statistics article “50 Years of Data Science” that, “[T]he activities that preoccupied [the statistics profession] over centuries are now in the limelight, but those activities are claimed to be bright shiny new and carried out by (although not actually invented by) upstarts and strangers. Various professional statistics organizations are reacting …” He cites three commentaries (Marie Davidian in Amstat News, July 2013; Bin Yu in IMS Bulletin, October 2014; and Marvin Goodson at the meeting of the Royal Statistical Society in May 2015). Three leaders in three leading societies in three subsequent years react to data science initiatives—long after these initiatives had taken hold in universities, government agencies, and corporations.

What are the messages to our field? Shall we avoid challenges and stay within our comfortable territory where our familiar approaches can be applied? (Recall the Tukey quotation at the end of my last column: “To be spared the responsibility of working on any of [these major problems] would make anyone’s life simpler and more pleasant.”) Or do we accept what Tukey called the “necessarily approximate nature of useful results” and that “indication procedures [may need] to grow up before the corresponding conclusion procedures”? In short, do we, as Tukey put it, “do what [we] can to clarify the issues and offer good advice … give up the protection of unquestioned hypotheses and contribute [our] best acuteness and wisdom under uncomfortable circumstances”?

Tukey advised us in The Future of Data Analysis that “data analysis is intrinsically an empirical science” and that, among other “necessary attitudes” for effective data analysis, we need to address “more realistic problems,” accept the “necessarily approximate nature of useful results,” and adopt the “free use of ad hoc and informal procedures in seeking indications.”

Panelists at JSM 2018 in Vancouver (Session 149: Theory versus Practice) made some of the same comments: We can justifiably take pride in demonstrating the mathematical properties of proposed statistical approaches, but perhaps we also need a balance between the time to demonstrate mathematical validity and the potential delay in announcing our breakthroughs while we wait for the “corresponding conclusion procedures” to “grow up.” [Recall that the statistics community was introduced to the jackknife via an abstract, seven sentences long, in The Annals of Mathematical Statistics.]

Can we find morals in these stories?

• We need to showcase all our talents—logical thinking, identification of process steps, design of relevant data collection, analysis and inference, characterization of uncertainty, clear results.
• We need to not just seize opportunities, but create invitations to the table. Better yet, we need to create the demands for our talents.
• Sometimes we need to be prepared to use our skills to present reasonable approaches to solve problems, even before the theory has been developed, before others offer less sensible approaches.

Dictionary.com defines proactive as “serving to prepare for, intervene in, or control an expected occurrence or situation” and reactive as “to act in response to an agent or influence; to respond to a stimulus in a particular manner.” Most ASA members are too busy to be proactive. We’re too busy reacting to our plates full of problems! That is why we have an ASA—to address issues of concern to the statistics profession that will ensure the relevance of our future.

Toward that end, David Williamson’s Impact Initiative is designed to address ways the ASA can inspire proactive activities. The committee he is chairing hopes to collect examples of fields or problems in which statistics had a critical impact (item (a) above) and, more importantly, identify whole areas where statisticians have the opportunity to proactively create the demand for statistical input, which will include analysis and inference of data, big and small (item (b) above). Please offer your suggestions.

Martin Luther King Jr. was celebrated last month with reminders of a quotation from Mahatma Ghandi: “Be the change you wish to see in the world.” What can the ASA do to proactively meet the next big scientific initiative? Thank you for helping us move our profession from “reactive” to “proactive” mode!