Moy Research Group
Join my research group! The Moy Research Group works on projects related to local politics, race, data science, and causal inference. I am looking to collaborate with interested parties across the social and computational sciences.
A limited number of paid positions will be available for graduate students interested in working on the development of original datasets. I am particularly looking for students with experience in R, web scraping, and GIS. If you are an undergraduate, we can collaborate on a Dean’s Undergraduate Research Fund Grant.
If you are interested in volunteering or collaborating, please send an email to bryant.moy@nyu.edu.
Upcoming or Ongoing Projects.
Mayoral Email Project.
Cleaning, compiling and parsing email records from an earlier experiment. We will utilize natural language processing and ego-centric networks to analyze the data.
RDD + Change Point Analysis.
Incorporating change point analysis as a tool to improve the credibility of regression discontinuity designs by assessing the ability to correctly identify theoretically-derived breakpoints in naturally noisy data. Provide a sequential workflow for applied researchers.
Measuring Police Culture.
Collect images and snapshots of local police department websites. We will use supervised and semi-supervised machine learning to create a novel measure of police culture.
Crime-Free Housing Ordinances.
Building a national dataset of crime-free housing ordinances. The project involves extensive data entry and web scraping. (Paid undergraduate RA positions available for qualitatively categorizing policies scraped from municipal websites.)
Local Campaign Websites.
Web-scrape local candidates’ campaign websites and archive a snapshot of the site, text, and images used. We will use this data to answer questions about presentation style and strategic behavior.
Neural Nets and Unmodeled Interactions in Binary Dependent Variable Models.
Unmodeled interactions in logistic regression render quantities of interests biased and inconsistent. Some argue that we should use fully moderated models, while other argue in favor of ML approaches (i.e., Post-Double Selection LASSO, Kernel-Regularized Least Squares, Bayesian Additive Regression Trees). We propose using Neural Networks to model interactions implicitly within hidden layers. (Collaboration potential with graduate students.)