Presentations

Susan Murphy presents "Assessing Personalization in Digital Health", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, December 1, 2021
Reinforcement Learning provides an attractive suite of online learning methods for personalizing interventions in a Digital Health. However after an reinforcement learning algorithm has been run in a clinical study, how do we assess whether personalization occurred? We might find users for whom it appears that the algorithm has indeed learned in which contexts the user is more responsive to a particular intervention. But could this have happened completely by chance? We discuss some first approaches to addressing these...
Read more about Susan Murphy presents "Assessing Personalization in Digital Health"
Xiang Zhou presents "Higher Education and the Black-White Earnings Gap", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, November 17, 2021
Higher education can be a double-edged sword in shaping the black-white earnings gap. It may serve as an equalizer, if black youth benefit more from attending and completing college than white youth. It may also reinforce racial inequality, given that black college-goers are less likely to complete college relative to white students. We employ a novel causal decomposition and a debiased machine learning method to isolate the equalizing and disequalizing effects of college and unveil the sources of these effects. Analyzing data from the NLSY97, we find that among men, a BA degree has a strong... Read more about Xiang Zhou presents "Higher Education and the Black-White Earnings Gap"
Isaiah Andrews presents "Inference on Winners", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, November 10, 2021

Many empirical questions concern target parameters selected through optimization. For example, researchers may be interested in the effectiveness of the best policy found in a randomized trial, or the best-performing investment strategy based on historical data. Such settings give rise to a winner’s curse, where conventional estimates are biased and conventional confidence intervals are unreliable. This paper develops optimal confidence intervals and median-unbiased estimators that are valid conditional on the target selected and so overcome this winner’s curse. If one requires validity...

Read more about Isaiah Andrews presents "Inference on Winners"
James Robins presents "Target Trials: Emulating RCTs using Observational Longitudinal Data", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, November 3, 2021

Target trials are RCTs one would like to conduct but cannot for ethical, financial, and/or logistical reasons. As a consequence, we must emulate such trials from observational data. A novel aspect of target trial methodology is that, for purposes of data analysis, each subject in the observational study is ‘enrolled’ in all target trials for which the subject is eligible, instead of a single trial. I will compare the strengths and weakness of the target trial...

Read more about James Robins presents "Target Trials: Emulating RCTs using Observational Longitudinal Data"
Eli Ben-Michael presents "Safe Policy Learning through Extrapolation: Application to Pre-trial Risk Assessment", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, October 20, 2021

Algorithmic recommendations and decisions have become ubiquitous in today's society. Many of these and other data-driven policies are based on known, deterministic rules to ensure their transparency and interpretability. This is especially true when such policies are used for public policy decision-making. For example, algorithmic pre-trial risk assessments, which serve as our motivating application, provide relatively simple, deterministic classification scores and recommendations to help judges make release decisions. Unfortunately, existing methods for policy...

Read more about Eli Ben-Michael presents "Safe Policy Learning through Extrapolation: Application to Pre-trial Risk Assessment"
Ruobin Gong presents "Towards Good Statistical Inference from Differentially Private Data", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, October 13, 2021
Differential privacy (DP) brings provability and transparency to statistical disclosure limitation. When data users migrate their analysis onto private data products, there is no guarantee that a statistical model, otherwise suitable for non-private data, can still produce trustworthy conclusions. This talk contemplates two challenges in drawing good statistical inference from private data. When the DP mechanism is transparent, I discuss how approximate computation techniques can be adapted to produce exact inference with respect to the joint specification of the intended model and the DP... Read more about Ruobin Gong presents "Towards Good Statistical Inference from Differentially Private Data"
Neil Shephard presents "When do common time series estimands have nonparametric causal meaning", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, October 6, 2021

The nonparametric potential outcome system provides a foundational framework for giving conditions under which common predictive time series statistical estimands, such as the impulse response function, generalized impulse response function, local projection and local projection instrument variables, have a nonparametric causal interpretation in terms of dynamic causal effects.

 

This is joint work with Ashesh Rambachan.

Cory McCartan presents "Measuring and Modeling Neighborhoods", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, September 29, 2021

More granular geographical data has allowed social scientists to probe how residential neighborhoods are formed and how they influence attitudes and behavior. To facilitate such studies, we develop a survey tool that allows respondents to draw their neighborhoods on a map. We propose a hierarchical Bayesian model that can be used to analyze which factors shape their neighborhoods. We have conducted a survey of registered voters in Miami, New York City, and Phoenix, and  find that across these cities, voters are more likely to include same-race and co-partisan census...

Read more about Cory McCartan presents "Measuring and Modeling Neighborhoods"
Sherry Zaks presents "Do we know it when we see it? Conceptualization and Cascading Bias in Identifying "Rebel-to-Party Transition"", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, September 22, 2021
Studies on rebel-to-party transition suggest that incorporating former-rebels into post-conflict politics creates a tenable path toward stability and democratization. Notwithstanding the salience of these results, the rebel-to-party literature is racked with an unacknowledged conceptual tension that simultaneously demands---and paves the way for---reconciliation. On the one hand, scholars exhibit remarkable convergence on both the core meaning and stakes of rebel-to-party transition. On the other hand, the literature reveals nearly as many different definitions of "rebel-to-party transition"... Read more about Sherry Zaks presents "Do we know it when we see it? Conceptualization and Cascading Bias in Identifying "Rebel-to-Party Transition""
Christopher Kenny presents "The Use of Differential Privacy for Census Data and its Impact on Redistricting: The Case of the 2020 U.S. Census", at CGIS Knafel Building (K354) - 12:10-1:30 pm, Wednesday, September 15, 2021
Census statistics play a key role in public policy decisions and social science research. Yet given the risk of revealing individual information, many statistical agencies are considering disclosure control methods based on differential privacy, which add noise to tabulated data. Unlike other applications of differential privacy, however, census statistics must be post-processed after noise injection to be usable. We study the impact of the US Census Bureau's latest Disclosure Avoidance System (DAS) on a major application of census statistics---the redrawing of electoral districts. We find... Read more about Christopher Kenny presents "The Use of Differential Privacy for Census Data and its Impact on Redistricting: The Case of the 2020 U.S. Census"

Pages