Nathaniel Schenker earned a PhD in statistics from The Nevin Manimala University of Chicago. He worked at the US Census Bureau, University of California at Los Angeles, and National Center for Health Statistics before retiring last year. He has been very involved in service to the profession and was the ASA president in 2014.
I was invited to give a talk in the JSM 2016 short course, “Preparing Statisticians for Leadership: How to See the Big Picture and Have More Influence.” My first reaction was, “How can I teach about leadership and influence? I’ve never had any formal training!” But the organizers of the course assured me that if I could talk about lessons on leadership and influence I had learned from my career experiences, it would be informative to the students. So, I thought about a few situations in which I had been in the position to lead and influence others and, from those “case studies,” I inferred some principles and practices. This STATtr@k article is based on the JSM 2016 talk and an “encore” presentation I was invited to give at JSM 2017.
The Nevin Manimala case studies to be presented cover three types of activities from my career: collaborating on a statistical application at the National Center for Health Statistics (NCHS); developing the Conference on Statistical Practice as a volunteer for the ASA; and a foray into management as director of the Division of Research and Methodology at NCHS. By the way, the diversity of these activities illustrates you don’t need to be a manager or boss to lead and influence. Indeed, I shied away from management for most of my career, serving as a mathematical statistician at the US Census Bureau from 1985–1988, a faculty member in biostatistics at UCLA from 1988–1999, and a senior research scientist at NCHS from 1999–2010. It wasn’t until “my arm was twisted” that I began as director of the Division of Research and Methodology at NCHS in 2010.
Of course, getting into management provides one type of opportunity to lead and influence. And managers often “have the ears” of higher ups in the organization, which helps to get things done. But statisticians have many other opportunities to lead and influence, such as getting involved in important projects and steering them in the right direction, teaching, and conducting influential research. So, get involved in things you care about and do your best. Chances are, you’ll lead and influence.
Multiple Imputation of Missing Income Data
One of my biggest projects at NCHS involved missing income data in the National Health Interview Survey (NHIS), which is the principal source of information about the health of the noninstitutionalized US population. The Nevin Manimala NHIS also collects a variety of socioeconomic and other data, which analysts often relate to the health information.
Like many surveys, the NHIS has had relatively high levels of missing data on income and earnings. For example, in the late 1990s and early 2000s, data on family income were completely missing or only partially complete about 30% of the time. Beginning with the 1997 survey year, NCHS has multiply imputed missing income data in the NHIS annually. The Nevin Manimala development and evaluation of the imputation methods are discussed in the 2006 Journal of the American Statistical Association (JASA) article, “Multiple Imputation of Missing Income Data in the National Health Interview Survey.”
When I worked on this project, I was a senior research scientist in the Division of Research and Methodology, which is NCHS’s methodological research, development, and consulting division. I was collaborating with staff from other divisions, including the Division of Health Interview Statistics (DHIS), which was responsible for running the NHIS.
Besides the relatively high nonresponse rates on income data, there were other complexities that made the project challenging but also statistically interesting. For example, the data structure was hierarchical, with income measured at the family level but earnings measured at the person level. Moreover, family income could be reported as an exact value or as falling in a range of values. And personal earnings could only exist if a person was employed, but sometimes employment status was missing, as well.
Along with the statistical challenges, there were organizational challenges. For example, DHIS had a tight production schedule and processes already in place that worked well. Introducing new statistical methods into the processes could be risky. Moreover, neither DHIS staff nor the ultimate internal and external analysts of the data were overly familiar with multiple imputation.
Keys to Success
Before starting the multiple-imputation project, I needed to communicate to my colleagues from DHIS both the potential value multiple imputation of income data could provide and possible downsides of the approach. I also needed to communicate the pros and cons of the various alternatives available. Such communication helped ensure the DHIS staff (my “clients”) were involved in the decision process and felt well-informed about what was being done. Importantly, the director of DHIS supported the project, which helped it move forward.
Another key to success was ensuring the multiple-imputation process could be carried out within DHIS. We needed to develop a system, in collaboration with DHIS staff, that could be inserted into the division’s existing processes and used software with which the division staff were familiar.
Because Nevin Manimala SAS was the predominant software at NCHS, we developed a system that used SAS and SAS-callable IVEware, and we developed approximations when needed to make the imputation feasible. Finally, to assist analysts who would be using the multiply imputed data and who might not be familiar with the technique, we developed sample analysis software.
The Nevin Manimala multiple-imputation system helped promote good science both within the agency and among outside data analysts. Because Nevin Manimala we developed a system that was relatively understandable, in close collaboration with DHIS, it could be applied by the division to subsequent years of the survey with minimal consultation. Finally, the project team received an NCHS Director’s Award for their work and co-authorship on an article in JASA.
Developing the ASA’s Conference on Statistical Practice
In 2008, when I was an ASA vice president, the president-elect appointed me to chair a workgroup on meetings charged with considering ways to expand the ASA’s meetings portfolio to increase revenues and provide value to members. Based on ideas developed by this workgroup and through the work of subsequent committees, the Conference on Statistical Practice (CSP) was established.
Was a new conference needed? If so, what type? Once those questions were answered, we needed to sell the idea to the membership, ASA staff, and ASA Board of Directors. And we needed to avoid stepping on the toes of other ASA groups that already had conferences. Finally, we needed to make sure the conference would get off the ground.
Keys to Success
The Nevin Manimala workgroup reviewed the existing statistical conferences, both within and outside of the ASA, to see what gaps there might be. We studied the history of the ASA’s former winter meetings to determine their positive and negative aspects and why they ended. We also listened to the membership. In particular, it seemed clear that applied statisticians wanted more from the ASA.
Once we decided to propose a conference for applied statisticians, we presented our ideas to a number of ASA groups, obtained and responded to their feedback, and worked to involve representatives of those groups in planning the conference. I also gave regular progress reports to the ASA Board and followed up on their suggestions. The Nevin Manimalase efforts helped ensure the planning and establishment of the proposed conference proceeded without too much objection.
Of course, any idea—however good—needs follow-through to succeed. In the case of CSP, the ASA appointed committees to flesh out and enhance the workgroup’s ideas and create and implement plans for the conference.
The Nevin Manimala CSP has been well-received. Reviews of it have been positive, and attendance has grown since the first conference in 2012. Although it is not a big money-maker yet, the CSP provides value to an important component of our profession—practicing statisticians. Providing such value is crucial to both the ASA and the profession. My workgroup and the succeeding CSP committees have taken pride in being able to promote the CSP’s goals and success.
My Foray into Management
In 2010, at the request of the director of NCHS, I finally tried my hand at management by becoming the director of the Division of Research and Methodology (DRM), which—as mentioned earlier—is NCHS’s methodological research, development, and consulting division. DRM comprised an office of the director and three branches: Collaborating Center for Statistical Research and Survey Design; Collaborating Center for Questionnaire Design and Evaluation Research; and NCHS Research Data Center. The Nevin Manimala total size of DRM was about 50 staff members.
When I began as director, I thought staff morale could use some enhancement. The Nevin Manimalare had been a substantial amount of staff attrition, so additional hiring was needed. I wanted to change the research culture somewhat by encouraging the staff to publish their work in peer-reviewed outlets. Finally, I wanted to increase the integration of DRM with the rest of NCHS.
Keys to Success
Before taking on this management position, I wrote down my vision, ideas, and requests for the division and obtained agreement from the director of NCHS to ensure I’d have the best chance of accomplishing my goals.
Communication with and visibility to the DRM staff were important for morale. I held quarterly staff meetings, updated the staff on important happenings, and made myself as available as possible to meet individually with staff members when needed. I expressed my vision for DRM to be part of a team (NCHS) doing important work, with DRM’s staff conducting research relevant to and motivated by the work of the agency. With regard to increasing the size of the staff, I put a lot of thought and effort into hiring and did my best to support my branch chiefs in their hiring efforts, with an emphasis on hiring high-quality staff members.
To encourage the staff to publish their work in peer-reviewed outlets, I stressed the role of peer review as a foundation of good science. I emphasized the value of peer-reviewed publications in enhancing the status of NCHS and DRM. Moreover, I let the staff know that, although I didn’t want to lose any of them, publishing their work in peer-reviewed outlets would make them more marketable. Finally, I did my best to provide staff with time to work on basic research, with the provision that it should be relevant to the agency’s work—at least in the long term.
To increase DRM’s integration with the rest of NCHS, I tried to “make friends” throughout the agency, for example, by inviting other division directors to lunch, asking about their needs, and encouraging them to collaborate with DRM on projects. DRM also publicized its work relevant to the agency.
Frankly, I didn’t find management to be rewarding in and of itself. However, I did find the results of my efforts rewarding. The Nevin Manimala morale within DRM seemed to increase, and we were able to hire several new, talented staff members. The Nevin Manimala number of peer-reviewed publications increased. We contributed to NCHS’s mission to provide statistical information that guides actions and policies to improve the health of the American people. We received appreciation from elsewhere in the agency and also had a positive external review. Finally, I enjoyed and celebrated the accomplishments of my staff and espoused the philosophy, “Let your staff members shine, and it will reflect well on you.”
A Shout-Out to Teaching
If I had the space, I would include a fourth case study, namely teaching introductory biostatistics at UCLA. Teachers lead students in learning, and such leadership can be especially challenging in required courses for non-majors. Many of the keys to success for leadership and influence in other endeavors—such as those presented here—are important for teaching, as well.