Home » Featured

Nan Laird Honored with International Prize in Statistics

1 April 2021 No Comment

Nomination Cites Transformative Contributions to Random Effects Modeling

    Louise Ryan, ACEMS Chief Investigator and Professor of Statistics, University of Technology Sydney

      Like most Australians, I’m used to taking phone calls and Zoom meetings at strange hours. But I was particularly delighted to learn on a recent early-morning call with Ron Wasserstein that our nomination of Nan Laird as the 2021 recipient of the International Prize in Statistics had been successful! Ron asked if I could do a short writeup about Nan for Amstat News, so I decided to simply give you a slightly shortened and adapted version of the nomination statement.

      Nan Laird

      Nan had just been promoted to associate professor in the department of biostatistics at the Harvard School of Public Health (HSPH) when I started there as a postdoctoral fellow in 1983. She was already well recognized as a powerhouse within the department and in the broader community. Because it was quite rare in those days to encounter such a prominent and successful female academic, I was somewhat awestruck and looked to her as a mentor and role model. A few years later, she became my department chair and boss. A few years after that, I became her department chair and boss! Although we have never written a paper together, I consider her to be a cherished colleague. Indeed, it was my pleasure and honor to interview her and write about her fascinating life in a 2015 Statistical Science article, “A Conversation with Nan Laird.”

      Like other intellectual giants, Nan has been extraordinarily productive throughout her career and made ground-breaking contributions to several ­distinct areas of statistical science. Indeed, Google Scholar reports more than 174,000 citations of her work. But because the International Prize in Statistics is awarded for a single work or body of work, it was necessary to pick a single area of focus.

      One natural possibility was her work on missing data and the EM algorithm, accomplished during her PhD years (her first published paper, “Maximum-Likelihood with Incomplete Data via the EM Algorithm,” published in the Journal of the Royal Statistical Society, Series B in 1977, has more than 61,000 citations). Her work on meta-analysis was also a contender (her paper, “Meta-Analysis in Clinical Trials,” published in the Journal of Controlled Clinical Trials in 1986, has more than 28,000 citations). Nan has also made seminal contributions in statistical genetics, especially family-based association studies.

      In the end, we decided to focus on her ground-breaking contributions to random effects ­modeling for longitudinal data analysis. Indeed, this was her research focus when I met her at HSPH and I have witnessed first-hand the transformative effect her work has had, not only on my own work but on the work of so many other methodologically-focused statisticians and in practical application over multiple domains.

      As described in my 2015 Statistical Science article, the HSPH Biostatistics Department was tiny when Nan started as an assistant professor in 1975, but started to expand under the transformative leadership of Fred Mosteller and, soon thereafter, Marvin Zelen. Both strove to create a stimulating intellectual environment with young, talented statisticians working on interesting collaborative projects. They also gave the time and space for creative exploration of statistical methods ideas and challenges that arose from those collaborations. This was an environment in which Nan thrived.

      A fateful encounter was with another young assistant professor in the department, James Ware, who arrived in 1979 to work on the Harvard Six Cities Study. This large longitudinal cohort followed children recruited from six US cities with the goal of studying the impact of air pollution on the development of their lung function. Nan’s unique background and perspectives, honed during her PhD with Arthur Dempster in the Harvard Statistics Department, led her quite naturally to formulate the problem in terms of each child having a unique intercept and slope, which were influenced by exposure to air pollution. By viewing these random intercepts and slopes as unobserved or missing data, she was able to apply the EM algorithm, thereby achieving a brilliant analytic strategy that made fitting random effects models computationally feasible. While random effects models had been around for some time and understood from a theoretical perspective, fitting them was extremely difficult, except in certain restrictive settings (e.g., balanced and complete designs where all subjects were measured at exactly the same set of times).

      The 1982 paper Nan and Jim published in Biometrics, “Random Effects Models for Longitudinal Data: An Overview of Recent Results,” has earned more than 9,200 citations and is still getting well over 400 citations a year, even today. Nan and Jim were savvy enough to recognize the importance of disseminating their work with user-friendly software and did so via a Fortran program called REML, which many people soon started to use.

      It is important to remember this was in the early ’80s, when statistical software packages were in their infancy. Indeed, when SAS released their PROC Mixed module in 1992, it is my understanding that it was largely based on the Laird-Ware algorithm. Modern day advances in computational power have made it possible to fit mixed effects models more quickly using other algorithms. But, back in the 80s and 90s, recognition that the EM algorithm could be used to fit broad classes of random effects models resulted in an explosion of activity, not only in terms of applications but also in terms of statistical methodology ­development.

      Over the next several decades, Nan guided numerous students and postdoctoral fellows to explore a wide range of topics derived from her foundational paper with Jim. Building on the foundations she helped them lay, an impressive number of her students and mentees have gone on to forge prestigious careers of their own. One of them, Garrett Fitzmaurice, was co-author with her and Jim on a highly cited monograph on longitudinal data analysis, Applied Longitudinal Analysis.

      It is no surprise Nan has been recognized with many prestigious awards and honors throughout her career. However, having recently retired from her academic career at Harvard, being selected as the third winner of the International Prize in Statistics seems like a fitting, crowning accolade for her. I also believe having a female winner of the prize is powerful and timely. As I indicated above, Nan made a strong impression on me as a young postdoc and I vividly recall how, through her example, she inspired me to think that perhaps I could also succeed in my career.

      Almost 40 years later, our field has come a long way, but there is still much to be done. If you were to ask a young person in our profession today to close their eyes and imagine a world-class statistician, how many will see a woman? The impact of awarding the International Prize in Statistics to Nan not only recognizes her ground-breaking contributions to statistical science, but it will help empower many young females in our profession and inspire others to join our profession.

      Editor’s Note: Marie Davidian, Garrett Fitzmaurice, Ross Prentice, Kathryn Roeder, and Scott Zeger helped draft the nomination and wrote letters of support.

      1 Star2 Stars3 Stars4 Stars5 Stars (No Ratings Yet)

      Comments are closed.