Collaboration with Statisticians

Danny Modlin earned his bachelor’s degree in mathematics from Elon University. After teaching middle- and high-school mathematics for six years, he earned his master’s degree in mathematics from The University of North Carolina at Wilmington and his master’s degree in statistics from North Carolina State University. Modlin is employed as a statistical training specialist at SAS Institute Inc. and working toward his PhD in statistics.

Researchers and statisticians exist in a symbiotic world of learning and discovery. To complete degree requirements, researchers are required to complete an introductory statistics course, as statisticians are required to take science courses. You, as a researcher, are an expert in your field, as we, as statisticians, are in ours. So how do we get experts with diverse specialties to find common ground so our team can quickly achieve a mutual understanding of the problem at hand and effectively pursue a solution?

Fortunately, as a statistics student at North Carolina State University (NCSU), I was exposed to an effective collaboration method thanks to NCSU’s statistical consulting course. Students and faculty across campus could connect to our departmental web page and request statistical assistance with their projects or research. Students in the consulting class then met with the requesting individuals and assessed their problems. To assist me in writing this article, I asked two NCSU faculty members—Jason Osborne and William Hunt—for their input about collaborations. The following are a few viewpoints I gained from them and the consulting course.

Don’t Wait Until the End to Ask for a Statistician’s Assistance

There are many who think they are fully capable of performing the statistical analysis needed to solve any of their problems with their brief introduction to statistics, and there are some statistical topics that researchers can tackle without issue. However, both Osborne and Hunt agree that a statistician should be involved as soon as possible.

“If an experiment is to be undertaken and the design is not simple, then the statistician should be included as early as possible,” said Osborne. Hunt stated, “The statistician [should be] trained to be a team member from the very beginning.”

Getting a statistician involved at the end of a project could result in, “I’m sorry, but the data you collected will not be able to answer your question(s) of interest.” The hassle of going back to square one with just weeks left before a deadline is an unwanted complication to a researcher. For this reason, statisticians should be part of your projects from the beginning. We can assist in designing the experiment and ensure the proper tests are performed under validated assumptions.

Communication Is Key

During typical daily interactions, most would agree that communication skills are something we all have the chance to practice. Whether it be telling a joke, catching up on what you did over the weekend, or giving directions to some location, the goal is to be able to convey your thoughts to someone else in such a way that they can understand. With all this practice, why is communication between researchers and statisticians sometimes difficult?

I posed the question, “What is the most important thing that a researcher could do that would ease their collaboration with statisticians?” to both NCSU faculty members. Osborne replied, “Communicate clearly. Explain things in basic terms. Explain the big picture first and then get into details about sampling or experimental designs.”

I would concur with this. I have had collaborations in which my customer’s first question was a design question. I first asked for an overview of the project/experiment and the question(s) to be answered. I would like to add that everyone involved needs to remain patient. You may have to explain a concept a couple of times, but that repetition will aid the statistician in their understanding of the problem and better enable them to solve your problem. A ‘good’ statistician, as I see it, is one who repeats back to the researcher, in his/her own words, their understanding of the problem. A ‘good’ researcher ensures that this statement conveys the problem accurately. Your patience will pay dividends later.

Are there specific concepts researchers need to know that would assist in collaboration? Hunt said, “It would be helpful if they had both introductory statistics and a course in the design of experiments.”

I definitely agree with the design of experiments suggestion. As a student at NCSU, I took our design of experiments course and was pleasantly surprised to see that the majority of students were not from the statistics department. When the semester began, I was unsure of how this balance would determine the focus of the course, but it made it one of the best I had the opportunity to take at NCSU. Each student benefited from others’ diverse backgrounds. From the statistics perspective, we saw real-world applications brought in by nonstatistics students for the experimental designs discussed. The nonstatistics students, beyond learning other experimental designs, were exposed to the statistical conversations concerning the problems they presented to the instructor. This immersion started the lines of communication between our groups.

Osborne answered the question by saying, “The concept of ‘repeated sampling’ can sometimes be elusive. I often ask how things would be different if the exact same experiment were repeated, controlling for as much variation as possible.”

For the researcher who may not be familiar with the concept of repeated sampling, I simply restate the need for a statistician to be involved as early in the problem as possible.

Statisticians Are Researchers, Too

Those who have studied statistics do not spend their days consumed by the calculations of probability density functions. We statisticians do have interests in science, too. For example, I have always had an interest in hurricanes, so much so that my PhD research involves hurricane modeling. My interactions with meteorological researchers have been much smoother than if a different statistician had been involved without some interest and basic knowledge of weather.

As a researcher, you may not have the good fortune of finding that ideal statistician who has an interest in and prior knowledge of your field, but do not let that stop you from finding a good communicator who is willing to listen and learn more about your subject matter.

Final Thoughts

Statisticians and researchers are two groups of people who need each other. Despite this need, communication between us is not always fluid. With a little work, and perhaps taking a few of the suggestions mentioned above, a more stable bridge can be built.