This column is written for statisticians with master’s degrees and highlights areas of employment that will benefit statisticians at the master’s level. Comments and suggestions should be sent to Megan Murphy, Amstat News managing editor, at megan@amstat.org.

Contributing Editor
Susan Duke is manager of benefit risk evaluation at GlaxoSmithKline (GSK). She has worked at GSK for slightly more than half of her career, initially as a respiratory project statistician and later as co-lead of GSK’s statistical graphics initiative. Previously, she worked in biotech, devices, and drug delivery at companies in the Bay Area; Seattle, Washington; and Ft. Collins, Colorado. She lives in Chapel Hill, North Carolina, and is involved in local community efforts in her free time.

When I think about where my journey with statistical graphics began, I remember a communications class I took as a beginning freshman at UW-Green Bay (known fondly back then as Eco-U). As an unvarnished beginning college student, I took to heart what the professor taught about communicating: My intent and action are not enough. There are so many places a message can become misunderstood along the way from my intent and action of communicating to the receipt of the message—from my lack of perspective to the recipients’ knowledge and interest (and vice versa) to various obstructions (e.g., noise and competing messages).

The second aha moment came as a college senior during a biometrics course, this time about the value of statistics to scientists. It was my second course in statistics, and the light bulb went on. I could see what a fantastic tool statistics was for scientists, to use math to quantify science, to put a probability on whether the topic of interest happened due to chance alone or something more.

The third influence on my journey was more protracted. I had just begun my master’s work at Louisiana State University, studying marine science. My advisor encouraged me to take as many statistics classes as I could because he said it would make me a better scientist (and give me an advantage in the job market). He mentioned that two of his prior students had achieved master’s degrees in both marine science and applied statistics. So, after taking some courses in the experimental statistics department, I decided to get double master’s degrees.

I found consulting under the tutelage of my statistics advisor, James Geaghan, to be especially useful. Working with graduate students and professors from other departments on the data questions for their research was a fantastic bridge between coursework and real-life research. I don’t recall making graphs per se, but I do remember drawing pictures—this helped them understand statistical concepts and helped me understand the science.

Who would have thought then that my freshman year communications interest would lead to expertise in statistical graphics design and the opportunity to encourage improvements in graphics software (i.e., software that creates publication-quality graphs, is compliant with regulatory requirements, and does it without a lot of fuss)? Fast forward 18 years. With 12 years as a biotech statistician and six years as a big pharma statistician at GSK under my belt, I was asked to co-develop a graphics software initiative I had proposed.

We can always do better, but there has been real progress on this journey. How did it happen? For one, we had strong support from senior leaders in statistics. And we listened to the rank-and-file statisticians and programmers, who told us the number-one issue with creating more and better graphs was time.

A symptom such as “I don’t have enough time” can mean a number of things. Are the priorities appropriate for encouraging graphs? Is it possible that beefing up graphic design skills would get key messages across more clearly? Do we have to fight with the available software for hours on end to get the graphs we need? Is the quality of the graphic sufficient for presentation and publication? Is there simply not enough time in the day because so many responsibilities are piled on biopharma statisticians?

As a co-leader of our graphics initiative, I took on this assignment like a research project and continuously educated myself. For graphics design, I had the pleasure of taking graphic design courses from well-known graphics experts: Edward Tufte, who taught me how to think graphically; Naomi Robbins, who taught me the value of designing collaboratively; and Frank Harrell, who taught me how statistical graphics fits within the practice of biostatistics. I had the pleasure of co-authoring an internal graphic design course, which gave me a better understanding of course participants’ initial views about graphics and, from their feedback, in what ways they found it helpful. To better understand the cultural/people aspects of where sticking points might lie and how to rectify them, I gained valuable insights from The Eighth Habit and The Tipping Point and from an internal course, titled “Balancing Innovation and Execution.”

Changing culture is the most challenging aspect of doing anything differently. In my experience, it surpasses any methodology or technology barrier. Humans are wired to keep doing things the same way—it’s in our genes and in our culture. It takes courage and perseverance with a strong dose of support to make things ‘tip’ in an organization.

Having spoken with many statisticians about graphics, it seems a common experience that most of us weren’t taught or encouraged to use graphics (or communications) as part of our statistical training. I was fortunate in my training to learn graphics and communication skills in my applied statistics program, but most of my training in this area came from consulting experience (beginning in my master’s program). So, making graphics tip for statisticians—considering how to design graphs in the same way we consider the clinical endpoints chosen and the methods to analyze them—takes more than good software.

At the end of the day, I think statistical graphics is important because it is a key communications vehicle, helping us truly understand the meaning of the data. It is beneficial for all concerned when internal clinicians, regulatory authorities, prescribing physicians, and patients can more transparently understand what a medication does. It is in the statistician’s hands to make it so.