Becoming an Environmental Statistician

Lance Waller is the Rollins Professor and Chair of the department of biostatistics and bioinformatics at Emory University’s Rollins School of Public Health. Before coming to Emory, he was an associate professor of biostatistics at the University of Minnesota and University of Alabama at Birmingham.

When I was eight or nine years old, I completed a school assignment wherein I used crayons to draw a picture of my future career. The picture shows me flat on the ground, hiding behind a bush, and watching a robin with the word “biologist” spelled out above. Many decades later, I now find most of my career has involved more time meeting with people or working on a computer than crouching behind bushes. Nevertheless, I still enjoy reading about, thinking about, and talking about the environment and ecology, especially their relationship to human and animal health. This interest, combined with my training in statistics, provided the framework for a career in environmental statistics, an area I thoroughly enjoy and feel is ripe with research and career opportunities for statisticians. For those considering a career in environmental statistics, I offer a few suggestions.

If you find the application interesting, the statistics will be more fun.

Environmental applications are what make statistics “environmental.” Such applications also often make the statistics challenging since most environmental applications do not fit neatly into an experimental setting, and even results from controlled experiments are used to infer patterns in an uncontrolled, dynamic, and noisy setting. As such, environmental applications offer a rich set of problems requiring clever application of existing tools and development of new tools to address problems particular to the application at hand.

If you find the application engaging (e.g., if you are really curious as to how air pollution changes over the course of a day, how a sea turtle chooses a nesting site, or how an environmental bacterium can infect a person), these challenges become a puzzle and you find yourself pondering both the application and your methods of analysis as you walk the dog, drive to work, or mow the lawn.

Puzzles are much more fun than calculations alone, and finding just the right calculations to solve a puzzle can be rewarding and fun. In general, if you find the work fun, you think about it more often and in more depth, and you do better work.

Early in my career, I was envious of statisticians who worked on data sets I found interesting (e.g., crater patterns on Venus, dive patterns of elephant seals). I wondered how they could be so lucky to have collaborators drop such beautiful data sets in their laps. Over time, I found that if I sought out researchers in areas I found interesting (e.g., disease ecology), I would find individuals with interesting questions and have access to precisely these interesting data sets (e.g., the spatial diffusion of raccoon rabies).

After a few conversations (and dedicated reading on my part), we could often find questions they wanted to answer from data with many interesting features requiring nonstandard statistical analysis. When I could answer an environmental question from their data using methods from my toolbox, an ongoing collaboration began.

Once you establish a reputation as someone who can answer interesting questions, the interesting questions start to come to you.

Let the application drive the methods, not the other way around.

Since the application makes statistics “environmental,” be sure your methods answer an environmental question as well as a statistical question. Too often, statisticians are tempted to use methods we are familiar with that don’t quite address the environmental question of interest. Rather than answering the question we should, we answer the question we can.

Adding temporal correlations to a problem simply because our dissertation involved temporal effects in another application may be interesting statistically, but may not address the environmental question of interest.

In collaborative work, it is essential to understand the language of your collaborators and to be able to converse fluently in it. Read their literature, listen to their seminars, and ask a lot of questions. Being able to point to related work in the environmental field will enhance your vocabulary, thinking, and statistics. Every analysis tells a story, and knowing the background helps you tell a more compelling story.

Specialize by diversifying.

Environmental statistics involves many sorts of statistical skills: longitudinal and spatial statistics, time series, missing data, measurement error, dynamic systems, and many more. Fill your methodological toolbox with the skills related to the questions you find most interesting. Learn the breadth and depth of data sources available to address these questions. Learn to identify the statistical questions underlying the related environmental questions and familiarize yourself with other areas of application addressing similar statistical questions.

Recognize that, as a statistician, you have the ability to translate a solution from one application area to another. This is a valuable skill and often enables bigger advances than small improvements to standard analytic procedures within a particular environmental field.

Write statistically and environmentally.

Keep your ideas clear and show a logical progression for both statistically and environmentally savvy readers. Write clearly for both types of reader. This is not always easy, but the challenge is minimized the more you (as a statistician) read the environmental literature. You find out what “counts” as evidence, what arguments are compelling, and what previous research sets the context for your results. I find it helpful to think of guiding the reader through my thought process, tables, maps, figures, and results in a manner that, when I reach a conclusion from the data, the reader thinks, “Well, of course.”

Living in two worlds.

In summary, a successful environmental statistician reads, writes, and thinks environmentally as well as statistically. A successful environmental statistician has a wide variety of tools available and a keen eye for identifying the right tool for the job, even if that tool has never been applied to an environmental application. The environmental statistician may be flat on the ground, peering through the brush of complicated data, but she or he is still looking for the robin.