Faculty Research Seminar Series

Edward Kennedy - Robust estimation and inference for the local instrumental variable curve

Sep 26, 2016 from 12:00 PM to 01:30 PM

Instrumental variables are commonly used to estimate effects of a treatment afflicted by unmeasured confounding, and in practice instrumental variables are often continuous (e.g., measures of distance, or treatment preference). However, available methods for continuous instrumental variables have important limitations: they either require restrictive parametric assumptions for identification, or else rely on modeling both the outcome and treatment process well (and require modeling effect modification by all adjustment covariates). In this work we develop robust semiparametric estimators of a "local" effect curve among compliers, i.e., the effect among those who would take treatment for instrument values above some threshold and not below. The proposed methods do not require parametric assumptions, incorporate information about the instrument mechanism, allow for flexible data-adaptive estimation of effect modification by covariate subsets, and are doubly robust (i.e., robust to misspecification of either the instrument or treatment/outcome processes). We discuss asymptotic properties under weak conditions, and use the methods to study infant mortality effects of neonatal intensive care units with high versus low technical capacity, using travel time as an instrument. - Read More

Jaideep Srivastava - Social Computing: A QCRI Perspective

Oct 03, 2016 from 12:00 PM to 01:30 PM

Social Computing is a rapidly emerging cross-disciplinary area at the intersection of social/behavioral sciences and computing. It encompasses a number of themes, including socio-technical systems, whose goal is to build systems that enable groups of people to interact for various purposes, computational social science, whose goal is to analyze data collected from such systems to study social/behavioral theories in a more nuanced manner, and social computing applications, whose goal is to develop novel solutions to specific societal problems that are enabled by such systems. At QCRI we are engaged in a social computing research agenda that addresses each of the themes above, focusing on the following application areas: response to humanitarian crises, social media and news, behavioral health and well-being, and urban computing. In addition to publishable research, we build systems, operate services, and sometimes even transition technologies into commercialization. Our partners in research include a number of UN agencies including UN-OCHA, UNESCO, and the World Bank; Doctors Without Borders, World Food Program and others. Our commercial partners include Al Jazeera, Doha News, and others. Our academic partners include MIT, Imperial College, Qatar University, Georgetown U. in Qatar, Weil-Cornell Medical College in Qatar, and Northwestern U. in Qatar. In this talk we will present an overview of social computing at QCRI, with specific examples to illustrate our activities. - Read More

Lawrence Wein - Data-driven Operations Research Analyses in the Humanitarian Sector

Oct 10, 2016 from 12:00 PM to 01:30 PM

We briefly discuss six projects in the humanitarian sector: (1) allocating food aid for undernutrutioned children using data from a randomized trial in sub-Saharan Africa, (2) allocating food aid for undernutritioned children using data from a nutrition project in Guatemala, (3) analyzing the nutrition-disease nexus in the case of malaria, (4) allocating aid for health interventions to minimize childhood mortality, (5) assessing the impact of U.S.'s failure to use cash-based food assistance on child mortality, and (6) deriving individualized biometric identification for India's universal identification program. - Read More

Alex Chouldechova - Fair prediction with disparate impact: A study of bias in recidivism prediction instruments

Oct 17, 2016 from 12:00 PM to 01:30 PM

Recidivism prediction instruments provide decision makers with an assessment of the likelihood that a criminal defendant will reoffend at a future point in time. While such instruments are gaining increasing popularity across the country, their use is attracting tremendous controversy. Much of the controversy concerns potential discriminatory bias in the risk assessments that are produced. This work discusses a predictive bias criterion originating in the field of educational and psychological testing that has recently been applied to assess the fairness of recidivism prediction instruments. We show that, when recidivism prevalence differs across groups, the constraints imposed by the predictive bias criterion force false negative and false positive rates to also differ. We then demonstrate how differences in error rates can lead to disparate impact under policies that assign stricter penalties to higher-risk individuals. - Read More