Hi! I am Sean Kent fourth-year Ph.D. student in Statistics at the University of Wisconsin - Madison.
I primarily study multiple-instance learning and dynamic treatment regimes through my work with Professor Menggang Yu.
I've also worked on projects in machine learning, image recognition, optimization, causal inference, and data visualization through my coursework and industry experience.
I will occasionally blog here about various side-projects, with a recent focus on data-science visualization. My spare time is often filled with cooking, playing volleyball, and checking out new restaurants and bars.
I have often thought about testing some of my craps theories. The folks at Best Craps Strategy provided the impetus to look deeper. On their site, they have a page on craps systems: detailed strategies usually involving multiple bets at different times. Since it is tricky to know how the combination of bets will affect wins and losses, you can’t just look up the odds for these systems. My simulator offers an opportunity for a detailed and rigorous analysis. [Read More...]
In March—which feels like forever ago—the spread of COVID-19 in the U.S. was undeniable and many were left wondering what to do. At that point, there was a lot of information circulating about COVID-19, but not a ton of centralized resources to understand the broad patterns. I wasn’t sure where to start, but I decided I wanted to help however I could. [Read More...]
This blog post is all about personal growth. Last year, I was playing around with some data I scraped from BeerAdvocate’s website. In particular, I was looking at their Top 250 Rated Beers list, which contains information on the beers that BeerAdvocate users rate highest. From that data, I created the visualization in the following tweet [Read More...]
rsmatch is an R package designed to perform Risk Set Matching. Risk set matching is useful for causal inference in longitudinal studies where subjects are treated at varying time points. The main idea is that treated subjects can match with anyone who hasn't yet been treated and those who never get treatment, but each subject can only be used in one pair. This creates a mixed-integer programming problem that we implement based on Li, Propert, and Rosenbaum (2001) "Balanced Risk Set Matching". We use this methodology to understand the effect of a wealth-shock on mortality in retired individuals.
I'm currently a fourth-year Ph.D. student in the Department of Statistics. In May 2020, I completed my M.S. in Statistics and received a letter of merit for outstanding performance on the Master's Exam on statistical collaboration. My research currently explores new theoretical extensions to multiple-instance learning with applications to breast cancer tumor detection under Professor Menggang Yu. I'm also interested in using causal inference within dynamic treatment regimes, creating effective data visualization to gain understanding, and applying other statistical machine learning methods. Outside of research, I enjoy building community in the Statistics Department through leadership roles in the Statistics Graduate Student Association.
I completed my Bachelor of Science in Mathematics, with a minor in actuarial science from Notre Dame in 2017, but that hardly tells the whole story. I took courses ranging from honors general education courses, including Philosophy, History, and English, to challenging elective courses like Machine Learning, graduate-level Econometrics, Investment Theory, and Topological Data Analysis. In my spare time, I was an active participant in my residence hall community—serving as its representative in the Student Government Senate and running the hall's pizza restaurant, Keough Kitch.
[Madison, WI] Research under the direction of Professor Menggang Yu on several projects in the area of multiple-instance learning, causal inference, and reinforcement learning for dynamic treatment regimes. Some projects are in collaboration with the Health Innovation Program, a healthcare research unit within the University of Wisconsin School of Medicine and Public Health.
[Skokie, IL] Continuing on my previous experience at Paynet , I independently constructed a model for business owner's compensation based on location, industry, and Bureau of Labor Statistics data. For the first time, I saw this model through the entire process, including exploratory data analysis, design, building in R and SQL, and testing. I further worked under Paynet's Director of Statistical Model Development & Chief Economist for additional ad hoc projects and review of his work.
[Skokie, IL] Working under Paynet's Lead Statistical Modeler and Senior Economist, my major contributions were twofold. First, I researched and explored the opportunity of machine learning models in the heavily regulated commercial lending space. This project involved understanding the theoretical framework, best practices, and practical implementations of several models—neural networks, random forests, gradient boosting, and elastic net—as well as testing these models in python as a comparison to one of Paynet's current products. This brought the entire modeling team up to speed on the most popular machine learning techniques and proved we could ensemble a few different models to quickly benchmark the performance of the interpretable models currently in use. The later portion of the summer was spent creating tools for analyzing past performance of our AbsolutePD model, both to serve as guidance for AbsolutePD v2 and set up a framework for evaluating future model performance. Near automation was achieved by integrating SQL code in an R script, allowing for an evaluation of 4 key metrics across hundreds of categories and cross-categories. The SQL/R integration that I figured out saved countless hours of manually running similar SQL scripts, and could speed up many other processes in the Analytics department.
[Notre Dame, IN] Tucked away in the first floor of Keough Hall lies a small yet bustling pizza restaurant fondly called Keough Kitch. My good friend and I took over ownership in 2015. Combining a data-driven approach to management and a revamped menu and marketing strategy, we pushed sales above all past records available, and created a strong presence in the residence hall community. This was all possible because of the excellent team we hired and managed, excel workbooks containing dashboards where we could see key performance indicators at a glance, and superb restaurant knowledge for our age. I primarily handled the design and upkeep of the excel documents, planning and execution of marketing promotions, and all financial aspects; however, as with any business ownership, many other responsibilities could be added to this list.
[Skokie, IL] In my first stint at Paynet, I came in with no professional experience, and a moderate understanding of excel. By the time I finished, I had made significant contributions to a project that predicts the probability of default for any industry and location combination, became comfortable coding in R and SQL, and developed an advanced understanding of Excel. This evolution was all possible due to a passion for learning and an excellent boss. He started me off reading an academic paper that underpinned the model he was building, and soon I was documenting and assuring the quality of his R code. Because of my familiarity with his code—and because my boss was swamped with work—I finished writing his code and did the majority of the analysis of results for our probability of default model, which was based on a hierarchical credibility framework. With the remaining time, I helped build automated excel spreadsheets for various deliverables and helped design and run SQL queries on Paynet's database of over 23 million small business loans.
Copyright 2017 Sean Kent All Rights Reserved | Design By W3layouts