Principal Components: Danielle Lyles on Data Science in Higher Ed

How does data science power the university student experience behind the scenes? Danielle Lyles, data and evaluation scientist at the University of Colorado Boulder, tells us about her work.


Danielle Lyles is a data and evaluation scientist at the University of Colorado Boulder, who works jointly in the Office of Data Analytics and the Office of Undergraduate Education. 

A former mathematics faculty member, Danielle joined us on the Data Science Mixer podcast to discuss how the university is using data science to fuel admissions; to inform strategies to support student success; and to address diversity, equity and inclusion in their programs.

Here are three “principal components” of what Danielle shared.


1. Higher education is ripe for the skillful use of data science.

Data science in higher ed is a wide-open space.

Our biggest challenge is that the university wasn't founded on gathering data and bringing it together to better understand itself and the students. So the data is often very siloed. There's often no documentation. … We're working on breaking down silos across the university, leading cross-functional projects, and being an objective referee between the groups. So, for example, there are often competing goals between admissions and student success. Admissions needs lots of students, but we want the right students so they can be successful. We just allow them to be data-driven, show them what we find, and work together with both of them. Data science in higher ed is a wide-open space with lots of opportunities, so it's super fun to be working on stuff like this.

2. Exploratory data analysis reveals important patterns in student attrition and behavior.

We're trying to build as diverse a class as we can.

We are definitely always thinking about diversity and inclusion. Everyone that I work with cares about students and all the types of students that there are. We're trying to build as diverse a class as we can. …  I've done a lot of exploratory data analysis. For example, I looked at attrition and graduation of students who didn't get into the college that they applied for. We say, "Hey, you could still come here, but you can be in our program for exploratory studies. And you can try to transfer into the college that you wanted. Or maybe we'll help you find something that you like better, that's better for you." … When do they leave? How many of them leave? Where do they go? Do they go to another university or community college? Does it seem like they just drop out of higher ed? Which subgroups have the highest rates of attrition? ... If you control for first-generation college student status, ethnicity doesn't have that much of an effect on retention. It's got the leadership very interested in what we are doing with first-generation students and how we can better support them. 


3. Topic modeling distilled students’ open-ended survey responses into insights that informed the university’s pandemic response.

I found the common themes and their relative proportions.

We had this new student survey, and almost all of the students answered it. One of the questions we asked was, "What is one thing CU Boulder can do to make you feel more comfortable coming in fall 2020?" I was asked to do topic modeling on that, to find the common themes and their relative proportions. That information helped the university determine what to do for fall 2020. The most common theme was they wanted masks, and they wanted hand sanitizer everywhere. They had all of these cool ideas about hand sanitizer stations everywhere and distancing in the classroom, and making sure that the rules would be enforced. That was the most common theme.



These interview responses have been lightly edited for length and clarity.

The podcast show notes and a full transcript are available on the Alteryx Community.