Statistics department to host free public lecture on college admissions ‘clustering’
The Penn State Department of Statistics will host a free public lecture presented by Professor David Hunter, titled “My Statistical Consulting Work Was Declared Unconstitutional—and I'm Partly Glad,” from 4 to 5 p.m. on Wednesday, Oct. 5, in Berg Auditorium, 100 Huck Life Sciences Building, Penn State University Park.
“It was pretty exciting when my run-of-the-mill consulting project working with Michigan’s admissions office became part of a major court case,” Hunter said. “I guess not many statisticians get to see their work mentioned by the chief justice of the United States. Later, I realized this experience can be seen as a useful anecdote about how we make decisions in a world increasingly awash in data.”
Abstract
Hunter’s almost entirely nontechnical talk will first introduce and discuss clustering, which is generally defined as organizing a collection of objects into categories based on a set of measurements.
Statisticians frequently adopt what is called a model-based approach to clustering, which subtly alters the notion of what clustering means. Hunter will illustrate this point of view using a simple example, based on a 1975 photo taken in front of Old Main, of what is called "unsupervised" clustering. He will also describe an example of "supervised" clustering, involving college admissions and a hotly contested legal issue that he worked on as a graduate student in the late 1990s.
The legal case ultimately led to a decision by the United States Supreme Court that, among other things, declared the model Hunter had constructed unconstitutional. But while Hunter said he is unhappy about the decision itself, the fact that the case could be argued at all has positive implications for our increasingly data-dependent society.
Hunter will conclude his talk with examples where obscurity in the model-building aspect of clustering can lead to unfortunate societal outcomes.
About Hunter
Penn State statistics professor David R. Hunter earned his doctorate in statistics from the University of Michigan in 1999, following a math degree from Princeton University in 1992 and two years teaching mathematics at a public high school in New Hampshire. He has been at Penn State since 1999, and previously served as head of the statistics department from 2012 to 2018.
Hunter has published widely on statistical models for networks and is a co-creator of the statnet suite of packages for network analysis in the programming language R. He co-coined the term "MM algorithms" and has written extensively on this and other EM-like algorithms. He has also extended the theory and computational practice of unsupervised clustering using nonparametric finite mixture models.
Hunter’s awards and other honors include being elected a fellow of the American Statistical Association and a member of the International Statistical Institute, named co-recipient of the Richards Software Award from the International Network for Social Networks Analysis, and appointed a faculty fellow of Penn State Teaching and Learning with Technology and research fellow of Le Studium, CNRS, France.