2022 ASA President Katherine Ensor led a JSM late-breaking session, titled “It Takes More Than a Village: Capacity Building Efforts for the Statistics Community,” urging the statistical and data science communities to propose mathematical sciences research institutes (MSRIs) in response to the National Science Foundation’s current quinquennial call from its Division of Mathematical Sciences.

Tian Zheng, chair of statistics at Columbia University and a member of Ensor’s NSF-engagement working group, organized and proposed the late-breaking session with help from David Banks, who was a leader of the Statistical and Applied Mathematical Sciences Institute—an MSRI that ran from 2002–2021 in Research Triangle Park, North Carolina—and participated on the panel.

The panel featured another SAMSI leader, a former leader of the Mathematical Biosciences Institute—an MSRI that ran from 2002–2020 in Columbus, Ohio—a former NSF Division of Mathematical Sciences director, and leaders from related NSF-funded institutes. The panelists highlighted the importance and urgency of a statistics-led MSRI and provided advice to potential proposers.

The Imperative of a Statistics-Led MSRI

Drawing on decades of experience leading institutes or related entities, panelists encouraged the statistics and data science communities to submit strong proposals for the MSRI solicitation and outlined the following four key points for why statistics-led institutes are critical at this time:

1. What statistics brings to science. First and foremost, the panelists emphasized the perspective, insight, and other scientific skills statistics brings to scientific problems. The importance of statistics—the science of learning from data—is magnified in the era of data science, machine learning, and AI because of its central role in all three. As noted in a 2023 ASA Board statement, “Framing questions statistically allows leveraging data resources to extract knowledge and obtain better answers.”

    
   Beyond the merits of statistics itself, Sastry Pantula of California State University at San Bernadino reminded the audience that “statistics is inherently an interdisciplinary science, so saying we need to lead an interdisciplinary institute is redundant.” He also urged the statistics community to organize themselves around MSRI proposals not only in recognition of the importance of statistics-led institutes for the benefit of our profession, but also as a driver of innovation in all science, technology, engineering, and mathematics.

2. Statistics community attributes synchronize with MSRI criteria. The panel highlighted how the strengths of statistics and the statistics community match many aspects of the MSRI solicitation. In addition to the already-noted interdisciplinary work of statisticians (cue John Tukey: “The best thing about being a statistician is that you get to play in everyone’s backyard.), statisticians already excel in being a national resource, not just geographically but in terms of being of service to the entire mathematical sciences community and beyond. They also excel in developing the next generation of statisticians and data scientists and in knowledge dissemination and outreach. In addition to playing a key role in scientific discoveries, statisticians are leaders in fundamental research that has a long-term impact within statistical and data sciences.

    
   Panel members spoke to other strengths the statistics and data science communities have that could provide an MSRI broader impact: the greater percentage of women earning statistics degrees than other quantitative, physical science, or engineering fields, particularly at the PhD level; the active mentoring community; and the efforts to build greater diversity in the field through StatFest, Infinite Possibilities, and the ASA JEDI group.

3. The challenges for bringing statistics to the science leadership table. Statistics-led MSRIs can operate in ways that are conducive to fostering statistical research and reflect the cultural differences in research processes, education, and training between mathematics and statistics. Kate Calder of The University of Texas at Austin shared the anecdote of a non-statistics-led institute at which she was expected to teach statistics to all student participants in a morning, whereas a topic such as dynamical systems would be given more airtime to bring participants up to speed.

    
   More broadly, panelists noted a lack of understanding or misperceptions of statistics outside the community. For example, statistics is often seen as a bag of tools rather than a scientific discipline and approach, which the ASA Committee on Funded Research addressed in their 2017 document focused on the biomedical research community, “Statistical Issues Seen in Non-Statistics Proposals.”

   In statistics-led MSRIs, statistics would be hard-wired at the leadership table, so programming could allow for maximal impact on the statistics field and in the scientific domains of focus.

   Helen Zhang of The University of Arizona pointed out that the University of Arizona TRIPODS program she led raised the visibility of statistics at UA substantially and brought her into meetings with deans, allowing her to integrate statistics into the UA research culture firsthand. Zhang also noted her leadership of UA TRIPODS better enabled those in computer science, data science, mathematics, and statistics to work together effectively.

   Banks agreed on the difference in cultures, noting computer science has weekly meetings to keep work moving quickly and on track.

   Abel Rodriguez of the University of Washington emphasized the importance of statistics-led MSRIs to counter the heavy bias in data science toward the computer science community. Rodriguez joined Zhang and Banks in noting the very different approaches of the statistics and computer science communities, the former taking a more case-study approach and the latter taking a systems approach.

4. SAMSI and related institutes speak to statistics-led MSRI imperative. The experiences of those involved in SAMSI, TRIPODS, and related efforts were often referenced when speaking to the push for a statistics-led MSRI. When asked about the legacy of SAMSI, Banks and Richard L. Smith of The University of North Carolina spoke to their diverse and deep impacts. Banks highlighted SAMSI’s high-impact programs that led to substantial contributions to science such as astrostatistics, uncertainty quantification, and adversarial risk analysis. Smith spotlighted work in environmental statistics, forensics statistics, and ecology. He also noted the ‘long-tail’ impact of SAMSI, as the papers published in the years after a SAMSI workshop or sponsorship show. Additionally, both Banks and Smith noted the many graduate students and postdocs involved with SAMSI that have gone on to successful careers.

    
   Pantula gave his perspective as head of the North Carolina State University Department of Statistics from 2002–2010, sharing that the department benefited enormously from SAMSI’s affiliation with the university, as did the broader community. He noted SAMSI gave him an advantage when recruiting faculty, students, and postdocs and provided mentoring opportunities to fledgling statisticians across the United States. He also spoke about how SAMSI’s influence extended to the National Academies—particularly its Committee on Applied and Theoretical Statistics—and the Canadian Statistical Sciences Institute, as well as to scientists in India. Indeed, Smith was the first chair of the Canadian Statistical Sciences Institute’s governing board, tapped for his experience leading SAMSI, and thereby had a key role in selecting the institute’s first permanent director and eventual location in British Columbia.

   Rodriguez shared that the successful TRIPODS phase II program, for which he is a leader, borrowed heavily from the SAMSI model and experience. He also cited the influence of the 2019 Statistics at a Crossroads Conference and report in getting him to think beyond a disciplinary perspective.

Advice for Potential Proposers

Panelists were sanguine about the potential success of statistics-led MSRIs with the continued explosion of data, continued emergence of data science, recent excitement about machine learning and AI, and societal challenges in the data-heavy areas of climate change and public health. Their advice covered the gamut. Rodriguez recommended thinking about envisioned MSRIs as data institutes for which statistical questions are central but tackled in true collaboration with scientists in other disciplines. He also urged being strategic with the proposed advisory boards to ensure an effective multidisciplinary approach.

Besides drawing on lessons from SAMSI, TRIPODS, and other centers, panelists recommended revising the aforementioned Statistics at a Crossroads report and 2013 report of the London workshop on the future of statistics and science.

Some panelists urged a culture that would better facilitate interdisciplinary collaboration, such as generating ideas in quicker cycles, as done well by the computer science community.

Calder noted the opportunity to be creative with statistics-led MSRIs, for example, by making diversity-related objectives and programming central to the overall administration of the institute, rather than as a separate component as she has seen elsewhere. She also urged proposed MSRIs be accessible and inclusive in a way that is as transparent as possible about operations and getting involved (i.e., no unwritten rules).

Zhang suggested thinking carefully about cultivating multidisciplinary collaborations and not just those that are statistically focused. She noted how the computer science and math folks at the AI TRIPODS had collaborations with medicine and specific disciplines but not as much with each other because of the different cultures. The University of Arizona folks of various disciplines learned how to collaborate and were motivated by commonality and acknowledged differences.

Rodriguez suggested exploring with DMS geographically distributed leadership models like that of the Regional Innovation Engines or Pacific Institute for the Mathematical Sciences to broaden leadership and take advantage of strengths not necessarily located in one geographic area.

The following were among the many other recommendations made:

• Play to the strengths of the statistics and data science communities in terms of multidisciplinary research.
• Have a robust plan for involving a diverse pool of undergraduate students. At a minimum, promote workshops and programs through active outreach and sending invitations to students of all backgrounds (as opposed to passively announcing a workshop program with a registration link).
• Given the NSF’s new Technology, Innovation, and Partnership Directorate, identify and emphasize an impact component.
• Draw in all of the statistics and data science communities and relevant parts of the mathematics communities.
• Connect to data science, machine learning, and AI but don’t leave out mathematicians and other disciplines.
• Engage a professional proposal writer early.
• Consider a mechanism for supporting state and federal government in policy development (e.g., to study the impact of changing the age of retirement).
• Consider an international component.
• Have a sense of urgency and make use of ASA communities to collaborate.

If you are considering applying for an MSRI, contact ASA Director of Science Policy Steve Pierson.