Susan VanderPlas, graduate alumna of the statistics department, received the National Science Foundation (NSF) CAREER Award
for her research regarding integrated methods in using charts in realistic ways. She shared more about her research, time at Iowa State University, and future plans via written interview.
Will you please summarize, in a few sentences, the research you proposed to NSF and why it is important?
Historically, the ways researchers have measured the effectiveness of graphics using estimation accuracy (whether you can get a number out of the chart precisely) and reaction time (how quickly you can answer the question). Neither method reflects how we actually use charts — if accuracy is necessary, we use tables, and in a good chart, we can get lost looking for interesting findings and asking new questions. This project will develop new, integrated methods to measure how we use charts in more realistic ways, by examining not only estimation accuracy but also sense-making, the quality of predictions, and the ability to use charts to support real-world decision making. Better measurement methods will help researchers develop guidelines for making charts that reflect actual use and engagement, improving our ability to communicate science effectively.
In addition to your well-known work on statistical graphics, I see your most cited paper to date is on machine learning in forensics applications. Will you please share a few sentences about your past and/or current work in forensics and why it is important?
When I was last at Iowa State, I worked with the Center for Statistics and Applications in Forensic Evidence, designing algorithms for matching forensic pattern evidence. The problem is actually similar (almost the inverse) to understanding the perception of graphics — we have to understand what examiners are seeing in order to design numeric measures that replicate those features, and once we have that, we can use machine learning and statistics to create objective algorithms that have better accuracy, reproducibility, and repeatability than examiners subjective opinions. Across the country, examiners subjective opinions, which lack repeatability and reproducibility, are admitted into courts, though this is slowly changing - Oregon is the latest state to decide that subjective firearms and toolmark examination conclusions are not admissible in court. It is important to have good, objective methods ready to support forensic labs when courts decide that subjective methods are not permissible, and this is what I worked on when I was at CSAFE. I have continued to work on this problem, including working with lawyers to help them understand the statistical issues with subjective evaluation and writing legal briefs to submit to courts across the country. Statisticians can't just work on theorems - our discipline needs to be engaged in the public sphere, and we need to be working on problems that are critically important to people's lives. Forensics is one of many application areas that has the potential to have a huge impact on society.
You were an excellent graduate student at Iowa State, earning MS and PhD degrees in our program. What were the most valuable things you learned and experiences you had while a graduate student at Iowa State?
The most valuable things I learned at Iowa State concern the importance of computational reproducibility and the very applied data wrangling skills that I use every day. Stat 579 and Stat 585 were incredibly important courses in helping me to develop as a statistician, and I use the information I learned there every day. Participating in the Graphics group at ISU was also formative, and I am still attending Graphics group meetings, which have moved primarily online and now incorporate students and faculty from several universities.
You have had a very successful academic career so far. Would you like to share anything about your future plans?
At the moment, I am hoping to find a position at a university which values statistics and data science. Last semester, I fought incredibly hard to keep the Statistics department at UNL, but our efforts were not successful, and the department is being eliminated. So now it's time to move on and find a new program so that I can continue my research in graphics and forensics in a more supportive environment.