- The limits of nudges: Results of a statewide vaccination RCT
(With Nat Rabb, Megan Swindal, David Glick, Jake Bowers, Anna Tomasulo, Zayid Oyelami, David Yokum.)
Nature (in press).
An RCT on the effects of different messages encouraging people to get vaccinated. We find that several weeks after vaccinations were available to all study participants, no particular message has any significant effect on vaccine uptake.
- Early Alert Systems During a Pandemic: A Simulation Study on the Impact of Concept Drift
(With Paul Xu.)
Early Alert Systems are quickly becoming commonplace in the educational space, especially for predicting drop out. We explore potential effects of the COVID-19 pandemic on the quality of data these models receive and what the impacts on their predictions may be.
- Predictors of enrollment in opioid agonist therapy after opioid overdose or diagnosis with opioid use disorder: A cohort study
(With Alexandria Macmadu, Kimberly Paull, Rouba Youssef, Sivakumar Batthala, Elizabeth A. Samuels, Jesse L. Yedinak, Brandon D.L. Marshall.)
Drug and Alcohol Dependence. Volume 219, 1 February 2021.
Utilizing Rhode Island's Ecosystem of administrative data, we find that the majority (58%) of Medicaid recipients did not enroll in opioid agonist therapy within 6 months of diagnosis. We explore various potential protective factors.
- The number of quartic D4-fields ordered by conductor
(With Ali Altug, Arul Shankar, Ila Varma.)
Journal of the European Mathematical Society. Volume 23..
We exploit a symmetry of the group of symmetries of the square to give the asymptotic density of discriminants of D4-extensions of the rationals.
- A Cautionary Tail: A Framework and Case Study for Testing Predictive Model Validity
(With Peter C. Casey and David Yokum.)
MUD3 @ SIGKDD'18.
We attempt to find rats in the District of Columbia using machine learning and subject matter expertise from urban rodentology. We do well at predicting 311 calls, but much less well at extrapolation. We provide a framework for field studies for machine learning models.
- Using Machine Learning to Assess the Risk of and Prevent Water Main Breaks
(With Avishek Kumar, Ali Rizvi, Benjamin Brooks, R. Ali Vanderveld, Chad Kenney, Sam Edelstein, Adria Finch, Andrew Maxwell, Joe Zuckerbraun, Rayid Ghani.)
We build a model to predict which water mains will break in the coming year in Syracuse, NY. We perform quite well, and the model is being used to guide proactive water main repair.
- Estimating student proficiency: Deep is not the panacea
(With Xiong, X., Khajah, M., Lindsey, R.V., Zhao, S., Karklin, Y., Van Inwegen, E.G., Han, B., Ekanadham, C., Beck, J.E., Heffernan, N., Mozer, M.C..)
Workshop on Machine Learning in Education @ NIPS'16.
A summary of three distinct critiques of Deep Knowledge Tracing as a method of estimating student proficiency.
- Back to the basics: Bayesian extensions of IRT outperform neural networks for proficiency estimation
(With Yan Karklin, Bojian Han, and Chaitu Ekanadham.)
We show that the original Deep Knowledge Tracing paper is fundamentally flawed and that, after correcting its errors, that DKT is outperformed by a simple hierarchical model on multiple data sets.
- Combinatorics on Partial Word Correlations
(With Francine Blanchet-Sadri, Justin Fowler, Joshua D. Gafni.)
J. Combin. Theory Series A, Vol 117 Issue 6 (2010).
- Correlations of Partial Words
(With Francine Blanchet-Sadri, Joshua D. Gafni.)
- A Report on Felony Crime in the District of Columbia in 2016
As Office of the Deputy Mayor for Public Safety and Justice of the District of Columbia.
(With Eric Foster-Moore and Paul Testa.)
Council of the District of Columbia January 2018.
A summary of felony crimes (mandated by law) in the District of Columbia in 2016.
- Encouraging MPD Officer Applications via Postcard: A Randomized Controlled Trial
(With Don Braman, Rachel Breslin, Antonio Charland, Stacy Small, Cynthia Dumas, J. Weill, E. Linos, K. Reddy, and David Yokum.)
We performed an RCT to determine what direct mail outreach strategy is most effective for getting DC residents to sign up to be police officers.
- The Knewton Platform: A General-Purpose Adaptive Learning Platform
(With Zachary Nichols.)
A deep dive into how Knewton's online learning platform worked at the time.
- Reducing the Gap: How Adaptive Follow-Ups Help Struggling Students
(With Hillary Green-Lerman and David Kuntz.)
Online October 2015.
A retrospective study exploiting variation in who used the Knewton platform to determine the effects of immediate, targeted follow ups on proficiency.