Faculty Research

Jennifer Hill (co-director)


Dr. Hill works at the intersection of social policy research and methodological development. She is interested in methods and study designs that allow researchers to go beyond making purely associational observations to actually be able to answer causal questions. In particular she focuses on situations in which it is difficult or impossible to perform traditional randomized experiments, or when even seemingly pristine study designs are complicated by missing data or hierarchically structured data. Hill has published in a variety of leading journals including Journal of the American Statistical Association, American Political Science Review, American Journal of Public Health, and Developmental Psychology. Hill earned her PhD in Statistics at Harvard University in 2000 and completed a post-doctoral fellowship in Child and Family Policy at Columbia University's School of Social Work in 2002.

Hill's current methodological research focuses primarily on causal inference and missing data.

Causal Inference
Hill has been working for several years on methods that help to reduce our parametric assumptions in observation study settings where we want to make causal inferences. In the past few years she has started to focus more strongly on exploring and highlighting the advantages of using existing data mining tools to model the response surface (that is, the model of the outcome given the treatment variable and covariates) as an alternative to popular methods like propensity score matching and weighting. This has grown into work on determining when sufficient empirical counterfactuals exist to make inferences, accommodating missing covariate data and generalizing results from experiments to other samples or populations. A new line of research will focus on sensitivity analysis methods both for observational studies and randomized experiments with noncompliance.

Missing Data
Hill is currently working on resolving some of the outstanding issues in the world of multiple imputation. This is joint work with Andrew Gelman and a team of faculty members, post-docs and students at Columbia University. A primary focus of this research is development of a multiple imputation centers imputation package in R to address some of shortcomings of existing packages with the following goals: 1) developing more flexible imputation models so that imputations better reflect the true data structure, 2) creating diagnostics to assess both parametric and structural assumptions, 3) providing diagnostics to assess convergence, 4) making the software more responsive and user-friendly, 5) assessing model coherence in chained equation imputation models, 6) addressing common stumbling blocks such as perfect or near-perfect collinearity between predictors when there are many variables in the imputation model.

Marc Scott (co-director)


Dr. Scott's research involves the development of statistical models for longitudinal data, often associated with some aspect of the life course. In this setting, there are a variety of dependencies, which, when modeled appropriately, can yield insight into how trajectories unfold and distinguish themselves from one another. The standard toolbox of models for multilevel and panel data, while often adequate, sometimes fail to capture underlying heterogeneity, limiting the predictive or prognostic value of the findings. One setting in which extensions have proven useful is the identification of clusters, or patterns of common behavior. Scott has published in a variety of methodological and applied journals including Journal of the Royal Statistical Society, Journal of Educational and Behavioral Statistics, Journal of Labor Economics and Statistics in Medicine. Scott earned his PhD in Statistics at New York University in 1998 and spent two years, post-doctorate, working as a Senior Research Associate at the Institute on Education and the Economy at Teachers College, Columbia University, before coming to NYU in 2000.

Scott's methodological research focuses primarily on models for longitudinal data; in particular, mixture models and categorical data models.

Mixture models allow for multiple regimes in the mean and covariance structure, and each regime reflects a pattern that may inform hypothesis about the underlying process. Scott has developed mixture models for work and medical histories and is currently working on expanding this approach to the educational domain.

Modeling features of the life course trajectory, whether they be educational, economic or health-related, often involves information about nominal levels of a feature. While years of education may be measured on a scale, types of classes do not have a natural metric. While jobs are associated with earnings, the job types themselves are not easily compared (prestige measures being one attempt to ordinate different job classifications). The two strands of research Scott has explored in this arena are extensions of log-linear models for competing-risk event histories in which the number of competing states is large and clustering models for categorical data. The former rely on Bayesian computation techniques. The latter offer many methodological challenges, the first of which is establishing a metric upon which to base similarity of nominal components. He has been exploring models that establish a common metric empirically and applied these models to work and educational histories in collaboration with researchers at the University of Washington, Seattle, and Teachers College, Columbia University.

Jack Buckley

Dr. Buckley holds doctorate and master's degrees in political science with an emphasis on public policy and statistical methodology from SUNY at Stony Brook.While serving as as the Deputy Commissioner of the National Center for Education Statistics, he taught statistics and education policy as an adjunct assistant professor at Georgetown University. He has also held academic positions at Boston College and the State University of New York (SUNY) at Stony Brook. He has been an affiliated researcher with the National Center for the Study of Privatization of Education at Teachers College, Columbia University.  After several years of teaching at NYU, Dr. Buckley went on leave to serve as Commissioner of the National Center for Educational Statistics, the Federal statistical agency responsible for collecting data in all areas of education in the U.S.  He is currently Senior V.P., Research at the College Board.

Nicole Carnegie

Dr. Carnegie's research interests focus on the statistical analysis of social networks, mathematical modeling of infectious diseases, program evaluation methodology and applications for interventions in global health, poverty reduction and development. As part of her doctoral work at the University of Washington, she spent a number of years working with the UW network modeling group. Her recent work has focused on improving and comparing alternative methodologies for the estimation of the incidence of HIV infection.

Estimation of HIV incidence: Incidence is one of two primary measures of the current status of any epidemic, and good incidence measures are crucial for evaluating the impact of prevention efforts. In the case of HIV, this measure has proven to be difficult to estimate, in part due to the long-term nature of the disease. Dr. Carnegie's work in this area focuses on the two leading methodologies currently in use: testing for recent infection and estimation from serial prevalence data. The first is highly resource-intensive, but assumed to be more precise, whereas the second relies on data already routinely collected, but relies heavily on assumptions about the testing patterns of infected and uninfected individuals. Current work aims to both improve the precision of each estimator and to compare their performance in order to evaluate the supposed superiority of testing for recent infection.

Dynamic network modeling: Dr. Carnegie is working to develop tools for the estimation and simulation of dynamic networks from cross-sectional data. The models build on the cross-sectional ERGM framework used for analysis of static networks to allow for changes in relational status and entry/exit of nodes in the population. This work is joint with members of the Network Modeling Group at the University of Washington.

Dr. Carnegie is currently Assistant Professor of Biostatistics at the University of Wisconsin.

Peter Halpin


Peter Halpin is an applied statistician whose research focuses on psychometrics and educational measurement. His recent projects have addressed the role of test anxiety in high-stakes assessment and model specification issues concerning the use of continuous versus discrete latent variables. He is currently developing models for the time series analysis of computer-administered collaborative problem solving tasks. This research is at the interesection of collaborative learning, educational data mining, and psychometrics. Halpin has been published in Psychometrika, Structural Equation Modeling, and Multivariate Behavioral Research. He received his doctorate from Simon Fraser University in 2010, and held a Natural Science and Engineering Research Council of Canada postdoctoral fellowship at the University of Amsterdam through 2012.

Statistical modeling of incidental data from learning technologies presents some interesting opportunities. In context of collaboration, activity logs provide a time-stamped record of how students communicate and interact with one another while solving a complex problem. These time-stamped events can be modeled as a point process. In comparison with traditional psychometric models, which only condition on the ability of the single individual being assessed, point processes also condition the previous activities of all the other group members. Intuitively, this means that we reckon on the (demonstrated) ability of the entire group when estimating the (latent) ability of each specific individual. Point processes also serve to incorporate the timing of activities, whereas traditional models for assessment have only considered whether activities are "correct" or not. The development of appropriate psychometric models, algorithms for their estimation, and software for the assessment and data capture of collaborative problem solving tasks is the focal topic of this research.

Daphna Harel


Daphna Harel is an applied statistician who studies issues in psychometrics and educational measurement. Her work focuses on bridging practice and theory through the development of theoretically-justified guidelines and procedures. Dr. Harel earned her PhD from the Department of Mathematics and Statistics at McGill University.

Item Response Theory: Dr. Harel works on developing principled approaches to measurement problems in the social sciences. Harel advocates for the use of statistics and procedures that minimize the impact of misspecification between the data-generating model and the model used for analysis. Her work includes the development of a weighted summed score for ranking respondents based on responses to polytomous questionnaires and the impact of collapsing score categories.

Measurement in Medical Studies: Dr. Harel works on measurement of latent traits in chronically ill populations. Through her work with the Canadian Scleroderma Research Group, Harel has improved the quality of instruments used to measure health-related quality-of-life in patients with Systemic Sclerosis.

Ying Lu

Dr. Lu has an interdisciplinary educational background with a Ph.D. in Public Policy and Demography from Princeton University (2005) and a Ph.D. in Statistics from University of North Carolina at Chapel Hill (2009). Before joining NYU Steinhardt, Ying Lu was an Assistant Professor at the Departments of Sociology and Political Science at the University of Colorado at Boulder. She was also a faculty affiliate at the Institute of Behavioral Science at UCB.

Ying Lu is particularly interested in quantitative methodology in social and behavioral sciences, with applications in population and health. She has published statistical methodology papers in leading journals such as the Journal of the Royal Statistical Society, Statistical Science and Political Analysis. Her current research deals with statistical models for high dimensional data. In a paper that is under review she and a coauthor study model selection problem for linear mixed effect model where they propose apenalized least square method to simultaneously select and estimate the effective random components in mixed models. She is currently undertaking a couple of research projects that involve classification problems with high-dimension but low-information features.

Joel Middleton

Dr. Middleton's interests include design based estimation and causal inference in randomized experiments. He also studies voter behavior and political persuasion, and he has 10 years experience designing surveys and experimental interventions for political organizations. Dr. Middleton received his PhD in Political Science from Yale University, a Masters in Statistics from The George Washington University, a Masters in Psychology from Brown University, and his BS from Lewis and Clark College. 2012.

Dr. Middleton is currently Assistant Professor of Political Science, U.C./Berkeley.

Tod Mijanovich


Tod Mijanovich is a research practitioner in the field of impact evaluation of health and social service programs. He earned his doctorate at NYU's Wagner School, where he worked for several years studying health care utilization among disadvantaged populations, the impacts of comprehensive community initiatives, and neighborhood effects on health. His current work focuses on the use of administrative data to develop measures of health care quality, refine methods of medical risk adjustment, and predict future hospitalizations and health care costs. He is principal investigator or co-investigator on several ongoing program and policy evaluations, including evaluations of calorie labeling and other food policies, supportive housing for homeless adults, patient-centered medical homes, and care management for medically high-risk populations.

Sharon Weinberg


Sharon Weinberg holds an A.B. in mathematics and a Ph.D. in psychometrics and research design methodology from Cornell University. Her current research focuses on issues of equity in higher education and the application of quantitative methods to address them. A forthcoming research-based volume co-edited with NYU colleague, Lisa Stulberg, Diversity in American Higher Education: Toward A More Comprehensive Approach, is to be published by Routledge Press. The volume goes beyond the usual race and gender definition of diversity, covering a broad array of diversity issues in higher education, including, among others, sexual orientation, political conservatism, and policies and practices at the secondary school level and within the legal environment that impact diversity in higher education. Another publication, in 2010, "New Approaches for Addressing Two Key and Related Issues in Faculty Salaries: Compression and Cost of Living," addresses her interest in salary equity at the college level. Two more recent projects include exploring reasons that may serve to explain the recent female advantage in postsecondary enrollment, as well as the effect on higher education of the early 1990's uncapping of mandatory retirement ruling for college professors. Her book, Statistics Using SPSS: An Integrative Approach, co-authored with former graduate student, Sarah Abramowitz, was written, in large part, because of her interest in statistics education. It is now in its second edition (Cambridge University Press, 2008). She is the 2008 recipient of the NYU Steinhardt Award for Teaching Excellence.

Professor Weinberg has authored or co-authored more than sixty articles, books, and reports on statistical methodology, statistical education, evaluation, and on such applied areas as clinical and school psychology, special education, and higher education. She has published in such leading journals as Psychometrika, Multivariate Behavioral Research, the Journal of Educational and Behavioral Statistics, the Journal of Consulting and Clinical Psychology, the Learning Disability Quarterly, and the Journal of Higher Education. She is the recipient of several major grants from Federal Agencies, including the National Science Foundation, the National Institute of Drug Abuse, and the Office of Educational Research and Improvement. She is a former Vice Provost for Faculty Affairs at NYU, a member of the Editorial Advisory Board of the Educational Researcher, the official journal of the American Educational Research Association, and President of the Board of Directors of the Jewish Foundation for Education of Women (JFEW), which has an endowment of over 60 million dollars.