Analysis of home-field advantage in the NFL across time and space.
I am a Ph.D. student in the Department of Statistical Science at Duke University starting in September 2018. My research interests are in applications of statistics to complex data arising in social science, particularly those involving political polarization.
My work with my Ph.D. advisor Alexander Volfovsky and the Duke Polarization Lab has involved designing a large field experiment to reduce polarization by pairing Democrats and Republicans to have anonymous, online conversations. Current projects revolve around analyzing these conversations to identify what kind of conversation features cause depolarization, while disentangling causal interference from conversation partners and addressing the inherent high-dimensional nature of text data.
I also recently started a collaboration with Mike West and the company 84.51 to assist with retail forecasting at a large grocery store chain. This work involves developing methods to forecast jointly price, quantity, and revenue across 100s of products and multiple geographies based on planned discount strategies. Key methodological innovations focus on identifying when and how to use cross-product discount information (identifying complements and substitutes) and how to capture global (company-wide) and local (product- and store-specific) trends.
Prior to starting my Ph.D., I graduated from Carleton College with a B.A. in Economics and Statistics. I then worked as an analyst at Cornerstone Research for two years, helping to prepare expert testimony on civil litigation regarding financial regulation, anti-trust issues, and labor market discrimination. After that, I worked for Professor Steven Levitt as a Research Professional in the Becker Friedman Institute at the University of Chicago, assisting with research projects about early childhood education, campaign spending, and the psychology of perseverance.
PhD in Statistical Science, Expected 2023
Duke University
BA in Economics and Statistics, 2014
Carleton College
Myself and another Ph.D. student created a module to teach undergraduates statistical methods for text analysis focused on political applications. We created and taught variants for implementation in introductory and advanced classes. This work was funded by the Duke Rhodes Information Initiative and is documented here: https://bigdata.duke.edu/projects/text-analysis-political-speech.
I will eventually move my old Wordpress site to this one, but for now you can find all of my old posts here.
Analysis of home-field advantage in the NFL across time and space.
When predicting event times in the future, you should include that information in the training data.
Some thoughts on how to build causal graphs and the role of privilege in taking down Theranos.
I review two applications of predictive modeling in mental health, one successful and the other not: identifying school shootings before they happen and prioritizing callers to a crisis hotline.