Michael Harding‘s path to graduate school didn’t follow a linear model. After earning mathematics and statistics degrees from Boston University, he landed at Lehigh Valley Health Network in Pennsylvania for a stint as a data scientist, in the midst of the COVID-19 pandemic. Building disease models to predict case surges, he found himself diving deeper into the research literature, an experience that sparked a realization: he wanted to be part of that conversation, too.
Now in his fourth year as a Statistics PhD student at UW–Madison, Harding has carved out a research niche studying how the costs of data collection should shape the design of machine learning systems. He has also won multiple teaching awards for his work teaching introductory statistics courses (while helping students overcome math anxiety), collaborated across departments through the Statistical Consulting Group on projects across fields, and discovered that statistics is fundamentally about connections—between datasets, disciplines, and people. “Statistics is inherently interdisciplinary,” he said.
We sat down with Harding to talk about finding fulfillment in teaching, his current research focus, and the rewards of applying statistics to elevate all disciplines.
How did your work on COVID modeling influence your decision to pursue a PhD?
I got a job as a data scientist at Lehigh Valley Health Network in their consumer insights and analytics department. The big project that I worked on was doing COVID case modeling, to predict when we were going to see influxes of cases in our service region and figure out where we need to have staff and beds available, for example.
I had a background in statistics, so I dove back into what people were doing in the research space and worked on implementing some of those models with the data that we had available. I really liked that project, especially the aspect where I was reading what people were doing on the cutting edge. That reaffirmed that I wanted to be part of the research conversation, and regardless of where I end up long-term, pursuing a PhD was the best logical step to get there.
How did you land on your current area of research, studying the costs of data collection in machine learning?
I came into grad school thinking broadly that machine learning is interesting, but that’s a huge umbrella. I took a bunch of classes and went to various reading groups and topics courses in the department to build a broad foundation and seek out the research problems I was most interested in pursuing. My advisor, Kirthi Kandasamy, and our research group, thinks a lot about data sharing settings and how you might create good incentives in a system where people are sharing data. The social and economic aspects of how those incentives can affect machine learning systems is something that I’ve more recently grown interested in. What are the costs of collecting and sharing data, and how are those incorporated into the models we use?
You’ve won multiple teaching awards during your time here. Why do you think your approach has resonated with students?
There’s a lot of math anxiety built up in many students over the years. As an instructor for the introductory statistics courses, including for life sciences and engineering students, I had a great opportunity to help patch that over, break down the anxiety they have about these concepts, and allow them to rebuild confidence from the ground up.
I sometimes see students get anxious about how to proceed when approaching big-picture problems that ask them to combine concepts from class. What I like to do in that case is take a step back from the larger problem and remember the individual steps that they have shown they understand and help guide them along the path to string those steps together toward a final answer. I always appreciate the moments when I get to see it click for students, whether they’re coming to office hours or the Gross Learning Center (GLC), initially struggling with material but leaving understanding more than they did coming in.
You also work as a member of the Statistical Consulting Group. Can you talk about a memorable project or experience there?
Last fall, I assisted Aishwarya Veerabahu with a project on native and invasive fungi. There were aspects of the design and data collection that informed a slightly more nuanced way of analyzing it, accounting for the areas where they collected samples. She wasn’t entirely sure how to build out the statistical methodology to answer her research questions, and that’s where I’m glad we could help.
Aishwarya had some stats background and knew some of the models the team should be running, but we really cleaned up the piece that helped translate data into science, solidifying her project. I ran into her on the bus a few weeks after we had stopped meeting, and she said to me, “Thank you for helping me improve and clean up my research. I feel like our work together has made me a better scientist.” It’s great to know when your work has made an impact in that way.
[Read more about Aishwarya Veerabahu’s research in The Conversation.]
How do you think about collaboration across fields?
Statistics is inherently interdisciplinary, so collaboration is vital. There’s data out there to be studied, no matter what questions you’re asking as a researcher. In the consulting context, people do a lot of hard work in different fields, collecting data and designing experiments with great scientific questions to ask, and they sometimes just don’t know how to actually use their data to answer those questions. That process of translation—from science to data and then turning data back into science—is where statistics is central.
Finally, what advice would you give someone considering a PhD in Statistics at UW–Madison?
You don’t need to feel like you know where you want to be after graduate school before deciding to apply. I didn’t even know that my current research area existed before I came to Madison. You should, however, be interested in doing research and getting in the weeds, doing the messy parts of the math and understanding the stats behind everything. And you should have a real hunger to learn. If you’re excited about those things, going for a PhD is a great opportunity.
It makes a big difference when you show up and are open and on the lookout for events, talks, or classes in related fields that expose you to a lot of different opportunities. UW–Madison is a great environment to do that. There is a lot of collaboration and intersection with related departments, even just in the realm of math, statistics, computer science, and engineering. A lot of people are affiliates across two, three, four departments. So there are plenty of chances for creative collaborations across the university.
Learn more about PhD program options in the Department of Statistics.
Learn more about Michael Harding’s research and teaching on his website.