Past Seminars (2008 - 2015)


The PRIISM seminar series consists of research seminars of interest to an applied statistics audience, from innovative applications of applied statistics to novel statistical theory and methodology. This page consists of an archive of past events from 2008 to 2015. Click here for information about the current seminar series or sign up for our mailing list to receive upcoming event reminders.

Past Events

Date & Time & Location
Speaker Name, Title and Affiliation Topic
11-12; Mar. 25, 20153rd Fl. Conf. Rm Kimball Matthew Steinberg (U. Penn.) 

Classroom Context and Observed Teacher Performance: What Do Teacher Observation Scores Really Measure?

11-12; Dec. 10, 2014, 3rd Fl. Conf. Rm Kimball Theo Damoulas (NYU/CUSP)

Mining NYPD’s 911 Call Data: Resource Allocation, Crimes, and Civic Engagement

12:15-1:45 pm; Oct. 29, 2014, Kaufman Mgmt Ctr (KMC) 4-80 Lixing Zhu (Hong Kong Baptist University) 

Joint Seminar with Stern/IOMS: Covariance Estimation for Factor Analysis using Pivotal Variables

 Mon. March 10, 2014, 4-6:30pm (talk from 4:30-5:30); Jurow Lect. Hall @ 100 Wash. Sq. E. Jennifer Hill 

Causal Inference and Data Science -- Why they need each other

11-12; Feb. 26, 2014, 3rd Fl. Conf. Rm Kimball Daphna Harel  Didactic Talk: An Introduction to Item Response Theory and Its Applications
12-1:15; Feb. 26, 2014, 3rd Fl. Conf. Rm Kimball Daphna Harel  Research Talk: The effect of collapsing categories on the estimation of the latent trait
10:45--11:45; Feb. 25, 2014, 3rd Fl. Conf. Rm Kimball Luke Keele Didactic Talk: Causal Mediation Analysis
12-1:15; Feb. 24, 2014, 3rd Fl. Conf. Rm Kimball Luke Keele Research Talk: Estimating Post-Treatment Effect Modification With Generalized Structural Mean Models
2-3; Feb. 13, 2014, 3rd Fl. Conf. Rm Kimball  Ivan Diaz Didactic Talk:  Definition and estimation of causal effects for continuous exposures
 11-12:15; Feb. 13, 2014, 3rd Fl. Conf. Rm Kimball  Ivan Diaz Research Talk: Definition and estimation of causal effects for continuous exposures: theory and applications
11-12; Nov. 21, 2013, 3rd Fl. Conf. Rm Kimball David Ong Income attraction: An online dating field experiment
11-12; Nov. 7, 2013, 3rd Fl. Conf. Rm Kimball Adam Glynn, Harvard Univ.  Front-door Difference-in-Differences Estimators: The Effects of Early In-person Voting on Turnout
11:30, Oct. 30, 2013, FORDHAM UNIV. Peter F. Halpin, New York University Fordham symposium: Modeling and Scoring Collaborative Problem Solving Tasks
11-12:15; Oct. 24, 2013, 3rd Fl. Conf. Rm Kimball

Vincent Dorie (IES Postdoctoral Fellow in the PRIISM Center)

Gaussian Processes for Causal Inference
 11-12:15; Oct. 17, 2013, 3rd Fl. Conf. Rm Kimball  Daphna Harel,  McGill Dept. of Mathematics & Statistics  The Inadequacy of the Summed Score (and How You Can Fix It!)
11-12:15; Sept. 19, 2013, 3rd Fl. Conf. Rm Kimball Dr. Nicole Carnegie,
Dept. of Biostatistics,
Harvard School of Public Health
Linkage of viral sequences among
HIV-infected village residents in Botswana: estimation of clustering rates in the presence of missing data
 Apr. 18, 2012
11:15am-12:15pm 
Juan Bello  Brown Bag talk: Information Extraction from Music Audio
 Apr. 4, 2012
11am-12pm 
Drew Conway  The impact of data science on the social sciences: perspective of a political scientist
Mar. 20, 2012
1pm-2pm 
Ji Seung Yang  Talk: Estimation of Contextual Effects through Multilevel Latent Variable Modeling with a Metropolis-Hastings Robbins-Monro Algorithm
Mar. 20, 2012
10:30am 
Preeti Raghavan & Ying Lu Brown Bag Discussion: Statistical modelling strategies for analyzing human movement data
Mar. 19th, 2012
1pm-2pm 
Ji Seung Yang  Talk: "An Introduction to Item Response Theory"
Mar. 6th, 2012
2:30pm-3:30pm
Peter Halpin, University of Amsterdam Talk: "Three perspectives on item response theory"
Mar. 5th, 2012
2:30pm-3:30pm
Peter Halpin, University of Amsterdam
Talk: "Point process models of human dynamics"
Feb. 15th, 2012
11am-12pm
Jay Verkuilen, CUNY/Grad Center Brown Bag seminar: Model Comparison is Judgment, Model Selection is Decision Making
Dec. 7th, 2011
11am-12pm
Cyrus Samii, NYU FAS Politics Dealing with Attrition in Randomized Experiments: Non-parametric andSemi-Parametric Approaches
Nov. 14th, 2011
11:30am-12:30pm
Jack Buckley, Associate Professor of Applied Statistics, NYU  on leave serving as Commissioner of NCES Brown Bag discussion of recent initiatives at the NCES.
Nov. 9th, 2011
11am-12pm
Eric Loken, Research Associate Professor Department of Human Development and Family Studies, Pennsylvania State University  Brown Bag Talk: The Psychometrics of College Testing: Why Don't We Practice What We Teach?
 Oct. 13th, 2011
1pm-3pm
Krista Gile (Department of Mathematics and Statistics University of Massachusetts/Amherst) Brown Bag Talk: An "introduction"
to Respondent Driven Sampling (RDS) methodology.

May 13, 2011
11:30am-6:45pm
TBA 

11th Annual Northeast Political Methodology Meeting "11th Annual Northeast Political Methodology Meeting"

"HOW DOES OBAMA MATCH-UP?  Counterfactuals and the Role of Obama's Race in 2008"

"Using Nonparametric Bayesian Modeling to Fight Terrorism" 

April 15, 2011
1:00-2:30pm
Tisch Bldg.
LC-21

Roderick J. Little, Department of Biostatistics, University of Michigan, and Associate Director for Research and Methodology, Bureau of the Census Subsample Ignorable Likelihood for Regression Analysis with Missing Data

*this event is co-sponsored with NYU's Economics (FAS), Information, Operations and Management Sciences (Stern), Politics and Sociology (both FAS) depts.
March 23rd, 10:45am, 246 Greene Street, Room 506W Pat Sharkey, Faculty of Arts and Sciences/Sociology, NYU Brown Bag Talk: Confronting selection into and out of social settings: Neighborhood change and children's economic outcomes
March 2, 2011
11am, 285 Mercer, Floor 3
Russ Steele, McGill University Brown Bag Talk: Modelling Birthweight in the Presence of Gestational Age
Measurement Error - A Semi-parametric Multiple Imputation Model
November 3rd, 2010
10:45AM-12PM
Professor Jack Buckley, Department of Applied Statistics, Social Science, and Humanities, NYU Steinhardt / PRIISM Center Didactic Talk: Using Multilevel Data to Control for Unobserved Confounders: Fixed and Random Effects Approaches
October 29, 2010
3:15-4:45pm
Guido Imbens, Department of Economics, Harvard University

 Methods Talk: An Empirical Model for Strategic Network Formation  *this event is co-sponsored with NYU's Economics dept.

October 27th, 2010
10:45am-12pm
Pat Shrout, New York University, Dept. of Psychology Brown Bag: Lagged Effects of Conflict in Intimate Couples on Same-day Closeness.
September 17, 2010
12pm-1pm
KMC 5-80
Jianqing Fan, Frederick L. Moore '18 Professor of Finance and Professor of Statistics, Princeton University Forecasting Large Panel Data with Penalized Least-Squares
*this event is co-sponsored with Stern's IOMS dept.
May 20th-21st, 2010
9AM-5PM
Various The 2010 Atlantic Causal Inference Conference

May 5th, 2010
10:45AM-12PM
246 Greene Street, 3rd Floor

Jennifer Hill An Introduction to Multiple Imputation

April 21th, 2010
10:50AM-12PM
285 Mercer Street, 3rd Floor

Nicole Carnegie Estimation of HIV Incidence With Testing For Recent Infection.
April 7th 2010
10:45AM-12PM
246 Greene Street, 3rd Floor
David Rindskopf Tentative talk topic:
Meta-analysis of single case designs: applying multilevel models to analyze small batches of moderate-length time-series data often found in behavioral research

March 31st, 2010
10:50AM-12PM
246 Greene Street, 3rd Floor

Jack Buckley Cross-National Response Styles in International Educational Assessments:
Evidence from PISA 2006
March 24th, 2010
10:50AM-12PM
246 Greene Street, 3rd Floor
Ying Lu Variable Selection For Linear Mixed Effect Models
February 12, 2010
11:30-1:00pm, Kaufman Management Center, Room 5-90
Mark S. Handcock, Professor of Statistics, UCLA Stern IOMS Statistics talk on responden-driven sampling.

February 11, 2010
12:00-1:30pm
19 University Place, 1st floor lecture hall

Mark S. Handcock, Professor of Statistics, UCLA Statistical Methods for Combining Survey and Population-level Data
December 9th, 2009
11am-12noon
Michael Sobel, Columbia University Fixed Effects Models in Causal Inference
November 18, 2009
11:00am-12:00pm
Statistics faculty from various departments. Meet and greet. PRIISM Coffee Hour #1
October 1, 2009
11:00am - 2:00pm
Dr. Michael Foster, Professor of Maternal and
Child Health in the School of Public Health, University of North Carolina, Chapel Hill
Does Special Education Actually Work?
May 5, 2009
4:15pm - 5:30pm
Dr. Michael Greenstone, 3M Professor of Environmental Economics
Department of Economics at the Massachusetts Institute of Technology;
Research Associate at the National Bureau of Economic Research (NBER);
Nonresident Senior Fellow at Brookings
Weather & Death in India: Mechanisms and Implications for Climate Change
February 12, 2009 Mark Hansen, Associate Professor of Statistics at UCLA
Design|Media Art and Electrical Engineering
Data analysis in an 'expanded field'
October 14, 2008
10AM-
Andrew Gelman, Professor, Statistics and Political Science at Columbia University. Red State, Blue State, Rich State, Poor State:
Why Americans Vote the Way They Do

WHO: Matthew Steinberg

WHAT: Classroom Context and Observed Teacher Performance: What Do Teacher Observation Scores Really Measure?

WHEN: March 25, 2015

WHERE: 246 Greene Street, 3rd floor conference room

ABSTRACT: As federal, state, and local policy reforms mandate the implementation of more rigorous teacher evaluation systems, measures of teacher performance are increasingly being used to support improvements in teacher effectiveness and inform decisions related to teacher retention. Observations of teachers’ classroom instruction take a central role in these systems, accounting for the majority of a teacher’s summative evaluation rating upon which accountability decisions are based. This study explores the extent to which classroom context influences measures of teacher performance based on classroom observation scores. Using data from the Measures of Effective Teaching (MET) study, we find that the context in which teachers work—most notably, the incoming academic performance of their students—plays a critical role in determining teachers’ measured performance, even after accounting for teachers’ endowed instructional abilities. The influence of student achievement on measured teacher performance is particularly salient for English Language Arts (ELA) instruction; for aspects of classroom practice that depend on a teacher’s interactions with her students; and for subject-specific teachers compared with their generalist counterparts. Further, evidence suggests that the intentional sorting of teachers to students has a significant influence on measured ELA (though not math) instruction. Implications for high-stakes teacher-accountability policies are discussed.

BIO:Dr. Steinberg is an Assistant Professor of Education, with appointments in the Education Policy and Teaching, Learning and Leadership Divisions. He is the Faculty Methodologist for the University of Pennsylvania IES Pre-Doctoral Training Program, as well as a Faculty Fellow with the University of Pennsylvania Institute for Urban Research and an Affiliated Researcher with the University of Chicago Consortium on Chicago School Research.

Back to the top of the page.


WHO: Lixing Zhu, Department of Mathematics/Hong Kong Baptist University

WHAT: Estimation for ultra-high dimensional factor models: a pivotal variable

detection-based approach

WHEN: Wednesday October 29, 2014

ABSTRACT: For a factor model, the involved covariance matrix often has no row sparse structure because the common factors may lead some variables to strongly associate with many others. Under the ultra-high dimensional paradigm, this feature causes existing methods for sparse covariance matrices in the literature to be not directly applicable. In this paper, for a general covariance matrix, a novel approach to detect these variables that are called the pivotal variables is suggested. Then, two-stage estimation procedures are proposed to handle ultra-high dimensionality in a factor model. In these procedures, pivotal variable detection is performed as a screening step and then existing approaches are applied to refine the working model. The estimation efficiency can be promoted under weaker assumptions on the model structure. Simulations are conducted to examine the performance of the new method.

Back to the top of the page.


WHO: Theo Damoulas (CUSP)

WHAT: Mining NYPD’s 911 Call Data: Resource Allocation, Crimes, and Civic Engagement

WHEN: Weds. Dec. 10, 2014, 11am-12pm

WHERE: Kimball 3rd Fl. Conference Room

ABSTRACT: NYPD’s 911 calls capture some of the most interesting urban activity in New York City such as serious crimes, family disputes, bombing attacks, natural disasters, and of course prank phone calls.In this talk I will describe research in progress conducted at the Center for Urban Science and Progress at NYU, in collaboration with NYPD. The work spans multiple areas of applied statistical interest such as sampling bias, time series analysis, and spatial statistics. The domain is very rich and offers many opportunities for research in core statistical and computational areas such as causal inference, search and pattern matching algorithms, evidence and data integration, ensemble models, and uncertainty quantification. At the same time there is great potential for positively impacting the quality of life of New Yorkers, and the day-to-day operation of NYPD.

Back to the top of the page.


WHO: Luke Keele

WHAT: Estimating Post-Treatment Effect Modification With Generalized Structural Mean Models

WHEN: Monday, February 24, 2014, 12-1:15pm

WHERE: 246 Greene St (Kimball), 3rd floor conference room 301W

ABSTRACT: In randomized controlled trials, the evaluation of an overall treatment effect is often followed by effect modification or subgroup analyses, where the possibility of a different magnitude or direction of effect for varying values of a covariate is explored. While studies of effect modification are typically restricted to pretreatment covariates, longitudinal experimental designs permit the examination of treatment effect modification by intermediate outcomes, where intermediates are measured after treatment but before the final outcome. We present a generalized structural mean model (GSMM) for analyzing treatment effect modification by post-treatment covariates. The model can accommodate post-treatment effect modification with both full compliance and noncompliance to assigned treatment status. The methods are evaluated using a simulation study that demonstrates that our approach retains unbiased estimation of effect modification by intermediate variables which are affected by treatment and also predict outcomes. We illustrate the method using a randomized trial designed to promote re-employment through teaching skills to enhance self-esteem and inoculate job seekers against setbacks in the job search process. Our analysis provides some evidence that the intervention was much less successful among subjects that displayed higher levels of depression at intermediate post-treatment waves of the study.

BIO: Dr. Keele received his PhD in Political Science from the University of North Carolina at Chapel Hill.

Back to the top of the page.


WHO: Luke Keele

WHAT: Didactic Talk: Causal Mediation Analysis

WHEN: Tuesday, February 25, 2014, 10:45-11:45am

WHERE: 246 Greene St (Kimball), 3rd floor conference room 301W

ABSTRACT: Causal analysis in the social sciences has largely focused on the estimation of treatment effects. Researchers often also seek to understand how a causal relationship arises. That is, they wish to know why a treatment works. In this talk, I introduce causal mediation analysis, a statistical framework for analyzing how a specific treatment changes an outcome. Using the potential outcomes framework, I outline both the counterfactual comparison implied by a causal mediation analysis and exactly what assumptions are sufficient for identifying causal mediation effects. I highlight that commonly used statistical methods for identifying causal mechanisms rely upon untestable assumptions and and may be inappropriate even under those assumptions. Casual mediation analysis is illustrated via an intervention study that seeks to understand whether single-sex classrooms improve academic performance.

Back to the top of the page.


WHO: Daphna Harel

WHAT: Research Talk: The Effect of Collapsing Categories on the Estimation of the Latent Trait

WHEN: Wednesday, February 26, 2014, 12-1:15pm

WHERE: 246 Greene St (Kimball), 3rd floor conference room 301W

ABSTRACT: Researchers often collapse categories of ordinal data out of convenience or in an attempt to improve model performance. Collapsing categories is quite common when fitting item response theory (IRT) when items are deemed to behave poorly. In this talk, I define the true model for the collapsed data both from a marginal and conditional perspective and develop a new paradigm for thinking about the problem of collapsing categories. I explore the issue of collapsing categories through the lens of model misspecification and explore the asymptotic behaviour of the parameter estimates from the misspecified model. I review and critique several current methods for deciding when to collapse categories and present simulation results on the effect of collapsing on the estimation of the latent trait.

BIO: Daphna Harel is a PhD candidate in Probability and Statistics at McGill University, and will be graduating in August of this year.

Back to the top of the page.


WHO: Daphna Harel

WHAT: Didactic Talk: An Introduction to Item Response Theory and Its Applications

WHEN: Thursday, February 27, 2014, 11am-12pm

WHEN: 246 Greene St (Kimball), 3rd floor conference room 301W

ABSTRACT: When a trait or construct cannot be measured directly, researchers often use multi-item questionnaires or tests to collect data that can provide insight about the underlying (or latent) trait. Item Response Theory (IRT) provides a class of statistical models that relate these observed responses to the latent trait allowing for inference to be made while still accounting for item-level characteristics. In this talk, I will introduce four commonly used IRT models: the Rasch model, the two-parameter model, the Partial Credit model and the Generalized Partial Credit model. My comparison will focus on the interpretation of and selection amongst these four models. One common use of IRT models is to determine whether an item functions the same for all types of people. This issue of Differential Item Functioning will be explored in the case of dichotomous items for both the Rasch model and two-parameter model. Lastly, three important summary statistics, the empirical Bayes estimator, the summed score and the weighted summed score will be presented and the use of each will be explained, specifically for the Partial Credit model and Generalized Partial Credit model.

Back to the top of the page.


WHO: Ivan Diaz

WHAT: Research Talk: Definition and estimation of causal effects for continuous exposures: theory and applications

WHEN: Thursday, February 13, 2014, 11am-12:15pm

WHERE: 246 Greene St (Kimball), 3rd floor conference room

ABSTRACT: The definition of a causal effect typically involves counterfactual variables resulting from interventions that modify the exposure of interest deterministically. However, this approach might yield infeasible interventions in some applications. A stochastic intervention generalizes the framework to define counterfactuals in which the post-intervention exposure is stochastic rather than deterministic. In this talk I will present a new approach to causal effects based on stochastic interventions, I will focus on an application of this methodology to the definition and estimation of the causal effect of a shift of a continuous exposure. This parameter is of general interest since it generalizes the interpretation of the coefficient in a main effects regression model to a nonparametric model. I will discuss two estimators of the causal effect: an M-estimator and a targeted minimum loss based estimator (TMLE), both of them efficient in the nonparametric model. I will discuss the methods in the context of an application to the evaluation of the effect of physical activity on all-cause mortality in the elderly.

BIO: Dr. Diaz received his PhD in Biostatistics from the School of Public Health, University of California at Berkeley, under the direction of Mark van der Laan and is completing a postdoc at Johns Hopkins this year.

Back to the top of the page.


WHO: Ivan Diaz

WHAT: Definition and estimation of causal effects for continuous exposures

WHEN: Thursday February 13, 2014, 2-3pm

WHERE: 246 Greene St (Kimball), 3rd floor conference room

ABSTRACT: In this talk I will discuss some important practical aspects of the definition and interpretation of potential (also called counterfactual) outcomes. These aspects must be considered with care when defining estimands in causal inference for observational studies. In particular, when working with continuous exposures, interventions that result in the usual potential outcomes are often inconceivable. As a consequence, the standard framework fails to provide relevant answers to scientific questions about interventions on the exposure. To solve this problem, I will present a proposal that defines counterfactual outcomes in terms of plausible interventions. I will define the causal effect of such interventions, and present an outcome regression estimator whose implementation is straightforward using existing regression software. The methods will be illustrated using an application to the evaluation of the effect of physical activity on all-cause mortality in the elderly.

Back to the top of the page.


WHO: David Ong (Assistant Professor of Economics at Peking University Business School)

WHAT: Income attraction: An online dating field experiment

WHEN: Thursday, November 21, 11am-12:15pm,

WHERE: 3rd floor conference room in Kimball Hall (246 Greene St)

ABSTRACT: Marriage rates have been decreasing in the US contemporaneously as women’s relative wages have been increasing. We found the opposite pattern in China. Prior empirical studies with US marriage data indicate that women marry up (and men marry down) economically. Furthermore, if the wife earns more, less happiness and greater strife are reported, the gender gap in housework increases, and they are more likely to divorce. However, these observational studies cannot identify whether these consequences were due to men’s preference for lower income women, or women’s preference for higher income men, or to other factors. We complement this literature by measuring income based attraction in a field experiment. We randomly assigned income levels to 360 unique artificial profiles on a major online dating website and recorded the incomes of nearly 4000 visits. We found that men of all income levels visited women’s profiles with different income levels at roughly equal rates. In contrast, women at all income levels visited men with higher income at higher rates, and surprisingly, these higher rates increased with the women’s own income. Men with the highest level of income got ten times more visits than the lowest. We discuss how the gender difference in “income attraction” might shed light on marriage and gender wage patterns, the wage premium for married men, and other stylized facts, e.g., why the gender gap in housework is higher for women who earn more than their husbands. This is the first field experimental study of gender differences in preferences for mate income.

Back to the top of the page.


WHO: Adam Glynn, Harvard University

WHAT: Front-door Difference-in-Differences Estimators: The Effects of Early In-person Voting on Turnout

WHEN: Thursday, Nov. 7, 2013, 11am-12:15pm

WHERE: 3rd floor conference room in Kimball Hall (246 Greene St)

ABSTRACT: In this talk, we develop front-door difference-in-differences estimators that utilize mechanistic information from post-treatment variables in addition to information from pre-treatment covariates. Even when the front-door criterion does not hold, these estimators allow the identification of causal effects by utilizing assumptions that are analogous to standard difference-in-differences assumptions. We also demonstrate that causal effects can be bounded by front-door and front-door difference-in-differences estimators under relaxed assumptions. We illustrate these points with an application to the effects of early in-person voting on turnout. Despite recent claims that early in-person voting had either an undetectable effect or a negative effect on turnout in 2008, we find evidence that early in-person voting had small positive effects on turnout in Florida in 2008. Moreover, we find evidence that early in-person voting disproportionately benefits African-American turnout.

Back to the top of the page.


WHO: Vincent Dorie, IES Postdoctoral Fellow, PRIISM Center

WHAT: Gaussian Processes for Causal Inference

WHEN: Thursday, October 24, 2013, 11am-12:15pm

WHERE: 3rd floor conference room in Kimball Hall (246 Greene St)

ABSTRACT: This brown bag talk will provide a mathematical and literature background for Gaussian Processes (GP) and discuss the use of GP in non-parametric modeling of the response surface for use in making straightforward causal comparisons. Additional topics include scalability, incorporating treatment levels as a spatial dimension, and the requirements for a fully-automated "black box" system for causal inference.

Back to the top of the page.


WHO: Nicole Carnegie, Harvard University

WHAT: Linkage of viral sequences among HIV-infected village residents in Botswana: estimation of clustering rates in the presence of missing data

WHEN: Thursday, Sept. 19, 2013, 11am-12:15pm

ABSTRACT: Linkage analysis is useful in investigating disease transmission dynamics and the effect of interventions on them, but estimates of probabilities of linkage between infected people from observed data can be biased downward when missingness is informative. We investigate variation in the rates at which subjects' viral genotypes link by viral load (low/high) and ART status using blood samples from household surveys in the Northeast sector of Mochudi, Botswana. The probability of obtaining a sequence from a sample varies with viral load; samples with low viral load are harder to amplify. Pairwise genetic distances were estimated from aligned nucleotide sequences of HIV-1C env gp120. It is first shown that the probability that randomly selected sequences are linked can be estimated consistently from observed data. This is then used to develop maximum likelihood estimates of the probability that a sequence from one group links to at least one sequence from another group under the assumption of independence across pairs. Furthermore, a resampling approach is developed that adjusts for the presence of correlation within individuals, with diagnostics for assessing the reliability of the method.

Sequences were obtained for 65% of subjects with high viral load (HVL, n=117), 54% of subjects with low viral load but not on ART (LVL, n=180), and 45% of subjects on ART (ART, n=126). The probability of linkage between two individuals is highest if both have HVL, and lowest if one has LVL and the other has LVL or is on ART. Linkage across groups is high for HVL and lower for LVL and ART. Adjustment for missing data increases the group-wise linkage rates by 40-100%, and changes the relative rates between groups. Bias in inferences regarding HIV viral linkage that arise from differential ability to genotype samples can be reduced by appropriate methods for accommodating missing data.

Back to the top of the page.


WHO: Daphna Harel, McGill University, Department of Mathematics and Statistics

WHAT: The Inadequacy of the Summed Score (and How You Can Fix It!)

WHEN: Thursday, October 17, 2013, 11am-12:15pm

WHERE: 3rd floor conference room in Kimball Hall (246 Greene St)

ABSTRACT: Health researchers often use patient and physician questionnaires to assess certain aspects of health status. Item Response Theory (IRT) provides a set of tools for examining the properties of the instrument and for estimation of the latent trait for each individual. In my research, I critically examine the usefulness of the summed score over items and an alternative weighted summed score (using weights computed from the IRT model) as an alternative to both the empirical Bayes estimator and maximum likelihood estimator for the Generalized Partial Credit Model. First, I will talk about two useful theoretical properties of the weighted summed score that I have proven as part of my work. Then I will relate the weighted summed score to other commonly used estimators of the latent trait. I will demonstrate the importance of these results in the context of both simulated and real data on the Center for Epidemiological Studies Depression Scale. 

Back to the top of the page.



WHO: 
Juan Bello, NYU
WHAT:  Brown Bag talk: Information Extraction from Music Audio
WHEN: Wed, April 18, 2012, 11:15am-12:15pm
WHERE: 246 Greene Street, Floor 3, Conference Room
ABSTRACT:  This talk will overview a mix of concepts, problems and techniques at the crossroads between signal processing, machine learning and music. I will start by motivating the use of content-based methods for the analysis and retrieval of music. Then, I will introduce work in three projects being investigated at the Music and Audio Research Lab (MARL): automatic chord recognition using hidden Markov models, music structure analysis using probabilistic latent component analysis, and feature learning using convolutional neural networks. In the process of doing so, I hope to illustrate some of the challenges and opportunities in the field of music informatics.

Read more about Professor Bello and the lab: 
https://files.nyu.edu/jb2843/public/ 
http://marl.smusic.nyu.edu/

Back to the top of the page.


 

WHO: Juan Bello, NYU
WHAT:  Brown Bag talk: Information Extraction from Music Audio
WHEN: Wed, April 18, 2012, 11:15am-12:15pm
WHERE: 246 Greene Street, Floor 3, Conference Room
ABSTRACT:  This talk will overview a mix of concepts, problems and techniques at the crossroads between signal processing, machine learning and music. I will start by motivating the use of content-based methods for the analysis and retrieval of music. Then, I will introduce work in three projects being investigated at the Music and Audio Research Lab (MARL): automatic chord recognition using hidden Markov models, music structure analysis using probabilistic latent component analysis, and feature learning using convolutional neural networks. In the process of doing so, I hope to illustrate some of the challenges and opportunities in the field of music informatics.

Read more about Professor Bello and the lab: 
https://files.nyu.edu/jb2843/public/ 
http://marl.smusic.nyu.edu/

Back to the top of the page.


WHO: Drew Conway
WHAT:  The impact of data science on the social sciences: perspective of a political scientist
WHEN: Wed, April 4, 2012, 11am-12pm
WHERE: 246 Greene Street, Floor 3, Conference Room
ABSTRACT:  As an emergent discipline, "data science" is by its very nature interdisciplinary.  But what separates this new discipline from traditional data mining work is a fundamental interest in human behavior.  Data science has been borne out of the proliferation of massive records of online human behavior, e.g., Facebook, Twitter, LinkedIn, etc.  It is the very presence of this data, and the accompanying tools for processing it, which have lead to the meteoric rise in demand for data science.  As such, principles from social science and a deep understanding of the data's substance represent core components in most data science endeavors.  In this talk I will describe this and the other core components of data science through examples from my own experience, highlighting the role of social science.


WHO: Ji Seung Yang 
WHAT: Talk: "Estimation of Contextual Effects through Multilevel Latent Variable Modeling with a Metropolis-Hastings Robbins-Monro Algorithm" 
WHEN: Tuesday, March 20, 2012, 1pm-2pm
WHERE: Pless Hall, 82 Washington Square East, 5th Floor Conference Room
ABSTRACT: Since human beings are social, their behaviors are naturally influenced by social groups such as one’s family, classroom, school, workplace, and country. Therefore, understanding human behaviors through not only an individual level perspective but also the lens of social context helps social researchers obtain a more complete picture of the individuals as well as society. The main theme of this talk is the definition and estimation of a contextual effect using nonlinear multilevel latent variable modeling  in which measurement error and sampling error are more properly addressed. The discussion is centered around an on-going research project that adopts a new algorithm, Metropolis-Hastings Robbins-Monro (MH-RM), to improve estimation efficiency in obtaining full-information maximum likelihood estimates (FIML) of the contextual effect. The MH-RM combines Markov chain Monte Carlo (MCMC) sampling and Stochastic Approximation to obtain FIML estimates more efficiently in complex models. This talk considers contextual effects not only as compositional effects but also as cross-level interactions, in which latent predictors are measured by categorical manifest variables. 

Back to the top of the page.


WHO: Preeti Raghavan and Ying Lu
WHAT: Brown Bag Discussion: Statistical modelling strategies for analyzing human movement data
WHEN: Tuesday, March 20, 2012, 10:30am-12pm
WHERE: 246 Greene Street, Floor 3, Conference Room
ABSTRACT:  Recent colloborations between Dr. Preeti Raghavan (Motor Recovery Lab, Rusk Institute) and Dr. Ying Lu (member of PRIISM) will be discussed in this talk. Using rich information of kinematic and EMG data collected at the Motor Recovery Lab, we are interested in the moverment patterns and how they change when the physiology is modified due to training, injury, disease and disability. We have explored Principle Component Analysis as a tool for dimension reduction to identify common patterns. Since the movement data are typically recorded over a period of time, it is important to model the movement pattern over time. We will discuss two aspects, treating the movement data as functional data (the functional approach) or as time series data. Accordingly we will discuss the use of functional PCA and dynamic factor analysis. Future directions of connecting EMG (muscle activities) with kinematic measures in these two contexts will also be discussed.

Back to the top of the page.


WHO: Ji Seung Yang
WHAT:  Talk: "An Introduction to Item Response Theory"
WHEN: Mon, March 19, 2012, 1pm – 2pm
WHERE: Pless Hall, 82 Washington Square East, 4th Floor, Payne Conference Room
ABSTRACT:  Item Response Theory (IRT) is a state-of-the-art method that has been widely used in large-scale educational assessments. Recently there has been an increased awareness of the potential benefits of IRT methodology not only in education but also in other fields such as health-related outcomes research and mental health assessment. This talk is to introduce fundamentals of IRT to an audience who is not acquainted with IRT. In addition to the key concepts of IRT, the three most popular IRT models for dichotomously scored responses will be illustrated, using an empirical data example extracted from Programme for International Student Assessment (PISA, OECD). This talk covers the principles of item analysis and scoring people in IRT framework and provides a list of advanced IRT topics at the end to sketch out the current methodological research stream in IRT. 

Back to the top of the page.


WHO: Peter Halpin, University of Amsterdam
WHAT:  Talk: "Three perspectives on item response theory"
WHEN: Tue, March 6, 2012, 2:30pm – 3:30pm
WHERE: Payne Room, 4th Floor, Pless Hall
ABSTRACT:  In this talk I introduce item response theory (IRT) to a general audience through consideration of three different perspectives. Firstly, I outline how IRT can be motivated with reference to classical test theory (CTT). This gives us the conventional view of IRT as a theory of test scores. Secondly, I compare IRT and discrete factor analysis (DFA). From a statistical perspective, the differences are largely a matter of emphasis. This situates IRT in the more general domain of latent variable modelling. Thirdly, I show how IRT can be represented in terms of generalized (non-) linear models. This leads to the notion of explanatory IRT, or the inclusion of covariates to model individual differences. Comparison of these perspectives allows for a relatively up-to-date “big picture” of IRT.

Back to the top of the page.


WHO: Peter Halpin, University of Amsterdam
WHAT:  Talk: "Point process models of human dynamics"
WHEN: Mon, March 5,  2012,2:30pm – 3:30pm
WHERE: Payne Room, 4th Floor, Pless Hall
ABSTRACT: There is an increasing demand for the analysis of intensive time series data collected on relatively few observational units. In this presentation I address the case of discrete events observed at irregular time points. In particular I discuss a class of models for coupled streams of events. These models have many natural applications in the study of human behaviour, of which I emphasize relationship counselling and classroom dynamics. I summarize my own results on parameter estimation and illustrate the model using an example from post graduate training. I also discuss ongoing developments regarding inclusion of random, time-varying covariates with measurement error and various other topics.

Back to the top of the page.


WHO:Jay Verkuilen (CUNY Graduate Center, Educational Psychology)
WHAT: Brown Bag Seminar: Model Comparison is Judgment, Model Selection is Decision Making
WHEN: February 15th, 2012, 11am-12pm
WHERE: 246 Greene Street, Floor 3, Conference Room
ABSTRACT: Model Comparison (MC) and Model Selection (MS) are now commonly used procedures in the statistical analysis of data in the behavioral and biological sciences. However, a number of puzzling questions seem to remain largely unexamined, many of which parallel issues that have been studied empirically in the judgment and decision making literature. In general, both MC and MS involve multiple criteria and are thus likely to be subject to the same difficulties as many other multi-criteria decision problems. For example, standard MS rules based upon Akaike weights employ a variation of Luce’s choice rule. The fact that Luce’s choice rule was constructed to encapsulate a probabilistic version of the ‘independence of irrelevant alternatives’ (IIA) condition has a number of consequences for the choice set of models to be compared. Contractions and dilations of the choice set are likely to be problematic, particularly given that information criteria measure only predictive success and not other aspects of the problem that are meaningful but more difficult to quantify, such as interpretability. In addition, in many models it is not entirely clear how to properly define quantities such as sample size or the number of parameters, and there are a number of key assumptions that are likely to be violated in common models, such as that of a regular likelihood. We consider some alternative ways of thinking about the problem. We offer some examples to illustrate, one using loglinear analysis and the other a binary mixed model.

Back to the top of the page.




WHO:
 Cyrus Samii (Department of Politics, NYU)
WHAT: Dealing with Attrition in Randomized Experiments: Non-parametric andSemi-Parametric Approaches
WHEN: December 7th, 201111am-12noon
WHERE: 
246 Greene Street, Floor 3, Conference Room
ABSTRACT: 
Uncontrolled missingness in experimental data may underminerandomization as the basis for unbiased inference of average treatmenteffects. This paper reviews methods that attempt to address thisproblem for inference on average treatment effects. I review inferencewith non-parametric bounds and inference with semi-parametricadjustment through inverse-probability weighting, imputation, andtheir combination.  The analysis is rooted in the Neyman-Rubinpotential outcomes model, which helps to expose key assumptionsnecessary for identification and also for valid statistical inference(e.g., interval construction).

Back to the top of the page.


 WHO: Eric Loken, Research Associate Professor Department of Human Development and
Family Studies, Pennsylvania State University 
WHAT: The Psychometrics of College Testing: Why Don't We Practice What We Teach?
WHEN: November 9th, 201111:30am-12:30pm
WHERE: 
246 Greene Street, Floor 3, Conference Room
ABSTRACT: 
Universities with large introductory classes are essentially operating like major testing organizations. The college assessment model, however, is many decades old, and almost no attention is given to evaluating the psychometric properties of classroom testing. This is surprising considering risks in accountability, and lost opportunities for innovation in pedagogy. As used in colleges, multiple choice tests are often guaranteed to provide unequal information across the ability spectrum, and almost nothing is known about the consistency of measurement properties across subgroups. Course management systems that encourage testing from item banks can expose students to dramatically unequal assessment. Aside from issues of fairness and validity, the neglect of research on testing in undergraduate classes represents a missed opportunity to take an empirical approach to pedagogy. Years of testing have generated vast amounts of data on student performance. These data can be leveraged to inform pedagological approaches. They can also be leveraged to provide novel assessments and tools to better encourage and measure student learning.

Back to the top of the page.



WHO: Krista Gile (Department of Mathematics and Statistics University of  Massachusetts/Amherst)
WHAT: An "introduction" to Respondent Driven Sampling (RDS) methodology.
WHEN: October 13th, 2011, 1pm-3pm
WHERE:
246 Greene Street, Floor 3, Conference Room
ABSTRACT:
Krista Gile (Department of Mathematics and Statistics University of Massachusetts/Amherst) is a statistician who works closely with social and behavioral scientists in the area of RDS. RDS is an innovative sampling technique for studying hidden and hard-to-reach populations for which no sampling frame can be obtained. RDS has been widely used to  sample populations at high risk of HIV infection and has also been used to survey undocumented workers and migrants.

In addition to providing an introduction to RDS for the PRIISM community, Krista will also be giving a statistical methodology talk at the NYU Stern/IOMS Dept. on Friday. Details are available here

Back to the top of the page.



WHO:
 Roderick J. Little, Department of Biostatistics, University of Michigan,
and Associate Director for Research and Methodology, Bureau of the Census
WHAT:
Subsample Ignorable Likelihood for Regression Analysis with Missing Data
WHEN: April 15, 2011, 1:00-2:30pm
WHERE:
Tisch Bldg. LC-21 
ABSTRACT: 
Two common approaches to regression with missing covariates are complete-case analysis (CC) and ignorable likelihood (IL) methods. We review these approaches, and propose a hybrid class, subsample ignorable likelihood (SSIL) methods, which applies an IL method to the subsample of observations that are complete on one set of variables, but possibly incomplete on others. Conditions on the missing data mechanism are presented under which SSIL gives consistent estimates, but both CC and IL are inconsistent. We motivate and apply the proposed method to data from National Health and Nutrition Examination Survey, and illustrate properties of the methods by simulation. Extensions to non-likelihood analyses are also mentioned. (Joint Work with Nanhua Zhang)

Back to the top of the page.


WHO: Pat Sharkey, NYU Sociology
WHAT:
Confronting selection into and out of social settings: Neighborhood change and children's economic outcomes
WHEN: Wednesday, March 23rd, 2011, 10:45am-12noon
WHERE:
Kimball Hall (246 Greene St) Room 506W
ABSTRACT: Selection bias continues to be a central methodological problem facing observational research estimating the effects of social settings on individuals. This article develops a method to estimate the impact of change in a particular social setting, the residential neighborhood, that is designed to address non-random selection into a neighborhood and non-random selection out of a neighborhood. Utilizing matching to confront selection into neighborhood environments and instrumental variables to confront selection out of changing neighborhoods, the method is applied to assess the effect of a decline in neighborhood concentrated disadvantage on the economic fortunes of African American children living within changing neighborhoods. Substantive findings indicate that a one standard deviation decline in concentrated disadvantage leads to increases in African American children's adult economic outcomes, but no effects on educational attainment or health.

Back to the top of the page.


WHO: Russ Steel, McGill University
WHAT:
Modelling Birthweight in the Presence of Gestational Age Measurement Error - A Semi-parametric Multiple Imputation Model
WHEN: March 2nd, 2011
WHERE:
285 Mercer Floor 3, Conference Room
ABSTRACT: 
Gestational age is an important variable in perinatal research, as it is a strong predictor of mortality and other adverse outcomes, and is also a component of measures of fetal growth. However, gestational ages measured using the date of the last menstrual period (LMP) are prone to substantial errors. These errors are apparent in most population-based data sources, which often show such implausible features as a bimodal distribution of birth weight at early preterm gestational ages (≤ 34 weeks) and constant or declining mean birth weight at postterm gestational ages (≥ 42 weeks). These features are likely consequences of errors in gestational age. Gestational age plays a critical role in measurement of outcome (preterm birth, small for gestational age) and is an important predictor of subsequent outcomes. It is important in the development of fetal growth standards. Therefore, accurate measurement of gestational age, or, failing that, a reasonable understanding of the structure of measurement error in the gestational age variable, is critical for perinatal research. In this talk, I will discuss the challenges in adjusting for gestational age measurement error via multiple imputation. In particular, I will emphasize the tension between flexibly modelling the distribution of birthweights within a gestational age and allowing for gestational age measurement error. I will also discuss strategies for incorporating prior information about the measurement error distribution and averaging over uncertainty in the distribution of the birthweights conditional on the true gestational age.

Back to the top of the page.


WHO: Professor Jack Buckley, Department of Applied Statistics, Social Science, and Humanities, NYU Steinhardt / PRIISM Center
WHAT:
Didactic Talk: Using Multilevel Data to Control for Unobserved Confounders: Fixed and Random Effects Approaches
WHEN: November 3rd, 2010, 10:45AM-12PM
WHERE: 246 Greene Street, 3rd Floor
ABSTRACTA didactic talk is a lecture on a topic of importance to applied researchers. The presentation will have a greater focus on either teaching the basic properties of a less familiar method or emphasizing aspects of a more familiar methodology that are essential to good practice. The presentation level should be appropriate for faculty working in the quantitative social, behavioral, policy and allied health sciences, as well as their advanced graduate students.
ADDITIONAL MATERIAL: 
References

Back to the top of the page.


WHO: Guido Imbens, Harvard University, Dept. of Economics
WHEN: Friday October 29th, 2010, 3:15am-4:45pm
WHAT: Methods lecture: An Empirical Model for Strategic Network Formation. This talk is co-sponsored with the NYU Department of Economics.
WHERE: 246 Greene Street,1st Floor Lounge, just south of Waverly.
TOPIC: Abstract: We develop and analyze a tractable empirical model for strategic network formation that can be estimated with data from a single network at a single point in time. We model the network formation as a sequential process where in each period a single randomly selected pair of agents has the opportunity to form a link. Conditional on such an opportunity, a link will be formed if both agents view the
link as beneficial to them.  They base their decision on their own characteristics, the characteristics of the potential partner, and on features of the current state of the network, such as whether the two potential partners already have friends in common.  A key assumption is that agents do not take into account possible future changes to the network.  This assumption avoids complications with the presence of multiple equilibria, and also greatly simplifies the computational burden of analyzing these models.  We use Bayesian markov-chain-monte-carlo methods to obtain draws from the posterior distribution of interest.  We apply our methods to a social network of 669 high school students, with, in average, 4.6 friends. We then use the model to evaluate the effect of an alternative assignment to classes on the topology of the network.

Paper: An Empirical Model for Strategic Network Formation 
This is joint work with Nicholas Christakis, James Fowler, and Karthik Kalyanaraman.

Back to the top of the page.


WHO: Pat Shrout, New York University, Dept. of Psychology
WHEN: Wednesday October 27th, 2010, 10:45am-12noon
WHAT: Brown Bag. Coffee will be provided. This will be an informal discussion of the methodology associated with a work in progress.
WHERE: 246 Greene Street, 3rd floor Conference Room, just south of Waverly.
TOPIC: Pat Shrout, New York University, Dept. of Psychology will present work-in-progress that examines lagged effects of conflict in intimate couples on same-day closeness. The data is derived from daily diaries, and as such is more intensive (dense) than traditional longitudinal data. Pat will discuss open issues arising in model selection, which highlight the tension between model choice, substantive questions, interpretation and causality.

Back to the top of the page.


WHO: Jianqing Fan, Frederick L. Moore '18 Professor of Finance and Professor of Statistics, Princeton University
WHAT: Statistics in Society lecture. Forecasting Large Panel Data with Penalized Least-Squares. This talk is also co-sponsored by the Stern IOMS-Statistics Group 
WHEN: September 17, 2010 12:00pm-1pm 
WHERE:
KMC 5-80
ABSTRACT:  Large Panel data arise from many diverse fields such as economics, finance, meteorology, energy demand management and ecology where spatial-temporal data are collected. Neighborhood correlations allow us to better forecast future outcomes, yet neighborhood selection becomes an important and challenging task. In this talk, we introduce the penalized least-squares to select the neighborhood variables that have an impact on the forecasting power. An iterative two-scale approach will be introduced. The inherent error (noise level) will also be estimated in the high-dimensional regression problems, which serves as the benchmark for forecasting errors. The techniques will be illustrated in forecasting the US house price indices at various Core Based Statistical Area (CBSA) levels.

Jianqing Fan is Frederick L. Moore'18 Professor of Finance and Director of Committee of Statistical Studies at Princeton University, past president of the Institute of Mathematical Statistics (2006-2009) and president of International Chinese Statistical Association. He has coauthored two highly-regarded books on "Local Polynomial Modeling" (1996) and "Nonlinear time series: Parametric and Nonparametric Methods" (2003) and authored or coauthored over 150 articles on computational biology, financial econometrics, semiparametric and non-parametric modeling, statistical learning, nonlinear time series, survival analysis, longitudinal data analysis, and other aspects of theoretical and methodological statistics. He has been consistently ranked as a top 10 highly-cited mathematical scientist since the existence of such a ranking. His published work has been recognized by The 2000 COPSS Presidents' Award, given annually to an outstanding statistician under age 40, the Humboldt Research Award for lifetime achievement in 2006, the Morningside Gold Medal of Applied Mathematics in 2007, Guggenheim Fellow in 2009, and the election to fellow of American Associations for Advancement of Science, Institute of Mathematical Statistics, and American Statistical Association.

Back to the top of the page.


WHO: Professor Jennifer Hill, Department of Applied Statistics, Social Science, and Humanities,
                NYU Steinhardt / PRIISM Center
WHAT: Didactic Talk: An introduction to multiple imputation: a more principled missing data solution
WHEN: May 5th, 2010, 10:45AM-12PM
WHERE: 246 Greene Street, 3rd Floor
ABSTRACT: A didactic talk is a lecture on a topic of importance to applied researchers. The presentation will have a greater focus on either teaching the basic properties of a less familiar method or emphasizing aspects of a more familiar methodology that are essential to good practice. The presentation level should be appropriate for faculty working in the quantitative social, behavioral, policy and allied health sciences, as well as their advanced graduate students.

Back to the top of the page.


WHO: Ying Lu, Assistant Professor of Applied Statistics, Department of Applied Statistics, Social Science, and Humanities, NYU Steinhardt / PRIISM Center
WHAT: Brown Bag Talk - Variable Selection For Linear Mixed Effect Models
WHEN: March 24th, 2010, 10:50AM-12PM
WHERE: 246 Greene Street, 3rd Floor
ABSTRACT: Mixed effect models are fundamental tools for the analysis of longitudinal data, panel data and cross-sectional data. They are widely used by various fields of social sciences, medical and biological sciences. However, the complex nature of these models has made variable selection and parameter estimation a challenging problem. In this paper, we propose a simple iterative procedure that estimates and selects fixed and random effects for linear mixed models. In particular, we propose to utilize the partial consistency property of the random effect coefficients and select groups of random effects simultaneously via a data-oriented penalty function (the smoothly clipped absolute deviation penalty function). We show that the proposed method is a consistent variable selection procedure and possesses the Oracle properties. Simulation studies and a real data analysis are also conducted to empirically examine the performance of this procedure.

Back to the top of the page.


WHO: Mark S. Handcock
WHAT: Statistical Methods for Sampling Hidden Networked Populations
WHEN: February 12, 2010, 11:30-12:30pm; Kaufman Management Center, 5-90 
ABSTRACT: Part of the Stern IOMS-Statistics Seminar Series, this talk will provide an overview of probability models and inferential methods for the analysis of data collected using Respondent Driven Sampling (RDS). RDS is an innovative sampling technique for studying hidden and hard-to-reach populations for which no sampling frame can be obtained. RDS has been widely used to sample populations at high risk of HIV infection and has also been used to survey undocumented workers and migrants. RDS solves the problem of sampling from hidden populations by replacing independent random sampling from a sampling frame by a referral chain of dependent observations: starting with a small group of seed respondents chosen by the researcher, the study participants themselves recruit additional survey respondents by referring their friends into the study. As an alternative to frame-based sampling, the chain-referral approach employed by RDS can be extremely successful as a means of recruiting respondents.

Current estimation relies on sampling weights estimated by treating the sampling process as a random walk on a graph, where the graph is the social network of relations among members of the target population.

These estimates are based on strong assumptions allowing the sample to be treated as a probability sample. In particular, the current estimator assumes a with-replacement sample or small sample fraction, while in practice samples are without-replacement, and often include a large fraction of the population. A large sample fraction, combined with different mean nodal degrees for infected and uninfected population members, induces substantial bias in the estimates. We introduce a new estimator which accounts for the without-replacement nature of the sampling process, and removes this bias. We then briefly introduce a further extension which uses a parametric model for the underlying social network to reduce the bias induced by the initial convenience sample.

This is joint work with Krista J. Gile, Nuffield College, Oxford. The research papers used as a basis for this talk, can be found at Ms. Gile's website regarding the following topics:
"Respondent-Driven Sampling: An Assessment of Current Methodology" (2010). Sociological Methodology forthcoming.
"Modeling Networks from Sampled Data" (2010). Annals of Applied Statistics forthcoming.

Back to the top of the page.


WHO: Mark S. Handcock, Department of Statistics, University of California - Los Angeles
WHAT: The fifth PRIISM-organized Statistics in Society lecture, this talk is also co-sponsored by the Stern IOMS-Statistics Group.
WHEN: February 11, 2010, 12:00-1:30pm; 19 University Place, 1st floor lecture hall 
ABSTRACT: In many situations information from a sample of individuals can be supplemented by information from population level data on the relationship of the explanatory variable with the dependent variables. Sources of population level data include a census, vital events registration systems and other governmental administrative record systems. They contain too few variables, however, to estimate demographically interesting models. Thus in a typical situation the estimation is done by using sample survey data alone, and the information from complete enumeration procedures is ignored. Sample survey data, however, are subjected to sampling error and bias due to non- response, whereas population level data are comparatively free of sampling error and typically less biased from the effects of non-response.

In this talk we will review statistical methods for the incorporation of population level information and show it can lead to statistically more accurate estimates and better inference. Population level information can be incorporated via constraints on functions of the model parameters. In general the constraints are non-linear, making the task of maximum likelihood estimation more difficult. We present an alternative approach exploiting the notion of an empirical likelihood.

We give an application to demographic hazard modeling by combining panel survey data with birth registration data to estimate annual birth probabilities by parity.

This is joint work with Sanjay Chaudhuri (National University of Singapore), and Michael S. Rendall (RAND Corporation). The research paper, "Generalised Linear Models Incorporating Population Level Information: An Empirical Likelihood Based Approach” (2008) (with Sanjay Chaudhuri and Michael S. Rendall). Journal of the Royal Statistical Society, B, 70, Part 2, pp. 311-328, was used as a basis for this talk.

Back to the top of the page.


WHO: Michael Sobel, Columbia University
WHAT:Fixed Effects Models in Causal Inference, a wWork-in-progress that clarifies the role of fixed effects models in causal inference.  He will make explicit the assumptions researchers implicitly make when using such models and what is actually being estimated
both of which are commonly misunderstood by those who use this strategy to identify causal effects. Coffee and some alternative beverages will be provided.
WHEN: December 9th, 2009, 12:00-1:30pm; 19 University Place, 1st floor lecture hall  
WHERE: We meet in the 3rd floor conference room in Kimball Hall, which is 246 Greene St., just south of Waverly.

Back to the top of the page.


WHO: Dr. Michael Foster, Professor of Maternal and Child Health in the School of Public Health, University of North Carolina, Chapel Hill.
WHAT: Dr. Foster will present the 4th Statistics in Society lecture, entitled: "Does Special Education Actually Work?" This talk will explore the efficacy of current special education policies while highlighting the role of new methods in causal inference in helping to answer it. Jointly sponsored by the Departments of Teaching and Learning and Applied Psychology, and by the Institute for Human Development and Social Change. The lecture will be followed immediately by a reception celebrating the official launch of the PRIISM Center.
WHEN: Thursday, October 1, 2009, 11:00 AM - 2:00 PM
WHERE: Room 900, Kimmel Center
ABSTRACT: This presentation assesses the effect of special education on school dropout (that is, the timing of a significant interruption in schooling) for children at risk for emotional and behavioral disorders (EBD). The analysis assesses the extent to which involvement in special education services raises the likelihood of an interruption in schooling in the presence of time-dependent confounding by aggression.  By using a child's observed school interruption time and history of special education and aggression, this strategy for assessing causal effects (which relies on g-estimation) relates the observed timing of school interruption to the counterfactual;  that is, what would have occurred had the child never been involved in special education. This analysis involves data on 1089 children collected by the Fast Track project. Subject to important assumptions, our results indicate that involvement in special education services reduces time to school
interruption by a factor of 0.64 to 0.93. In conclusion the effcacy of special education services is questionable which suggests that more research should be devoted to developing effective school-based interventions for children with emotional and behavioral problems.

Back to the top of the page.


WHO: Dr. Michael Greenstone
WHAT: Weather & Death in India: Mechanisms and Implications for Climate Change - This Event is Free and Open to the Public
WHEN: May 5, 2009 4:15pm - 5:30pm
WHERE: NYU Kimmel Center, Room 914 (9th Floor), 60 Washington Square South
ABSTRACT: Is climate change truly a matter of life and death? Join us as acclaimed economist Dr. Michael Greenstone discusses revelatory new research on the impact of variations in weather on well-being in India. The results indicate that high temperatures dramatically increase mortality rates; for example, 1 additional day with a mean temperature above 32° C, relative to a day in the 22° - 24° C range, increases the annual mortality rate by 0.9% in rural areas. This effect appears to be related to substantial reductions in the income of agricultural laborers due to these same hot days. Finally, the estimated temperature-mortality relationship and state of the art climate change projections reveal a substantial increase in mortality due to climate change, which greatly exceeds the expected impact in the US and other developed countries. Co-sponsored by the Global MPH program, the NYU Steinhardt School of Culture, Education and Human Development, and the NYU Environmental Studies program. Presented as part of the ongoing series Statistics in Society, organized by the Steinhardt PRIISM Center.

Michael Greenstone is the 3M Professor of Environmental Economics in the Department of Economics at the Massachusetts Institute of Technology. He also is a Research Associate at the National Bureau of Economic Research (NBER) and a Nonresident Senior Fellow at Brookings. His research is focused on estimating the costs and benefits of environmental quality. He has worked extensively on the Clean Air Act and examined its impacts on air quality, manufacturing activity, housing prices, and infant mortality to assess its costs and benefits. He is currently engaged in a large scale project to estimate the economic costs of climate change. Other current projects include examinations of: the benefits of the Superfund program; the economic and health impacts of indoor air pollution in Orissa, India; individual's revealed value of a statistical life; the impact of air pollution on infant mortality in developing countries; and the costs of biodiversity.

Greenstone is also interested in the consequences of government regulation, more generally. He is conducting or has conducted research on: the effects of federal antidiscrimination laws on black infant mortality rates; the impacts of mandated disclosure laws on equity markets; and the welfare consequences of state and local subsidies given to businesses that locate within their jurisdictions. Greenstone received a Ph.D. in economics from Princeton University and a BA in economics with High Honors from Swarthmore College.

Back to the top of the page.


WHO: Mark Hansen, UCLA
WHAT: Data analysis in an 'expanded field'
WHEN:Thursday February 12, 2009, 12:00-1:15,
WHERE: Warren Weaver Hall, Room 1302
ABSTRACT:


Photo courtesy of Ben Rubin, EAR Studio

The Center for the Promotion of Research Involving Innovative
Statistical Methodology (PRIISM) is delighted to announce the second
Statistics in Society lecture for the academic year. 


Mark Hansen, a UCLA statistician with joint appointments in Electrical Engineering and Design/Media Art, will be giving a talk that examines the interface between statistics, computing and society entitled "Data analysis in an 'expanded field' ". 

Hansen is perhaps best known locally for his work co-creating a current art installation, "Movable Type" in the New York Times Building here in manhattan. However his research reaches far beyond this realm drawing on fields as diverse as information theory, numerical analysis, computer science, and ecology.

For instance, Hansen currently serves as Co-PI for the Center for Embedded Networked Sensing or CENS, an NSF Science and Technology Center http://research.cens.ucla.edu/ ) that describes itself as "a major research enterprise focused on developing wireless sensing systems and applying this revolutionary technology to critical scientific and societal pursuits. In the same way that the development of the Internet transformed our ability to communicate, the ever
decreasing size and cost of computing components is setting the stage for detection, processing, and communication technology to be embedded throughout the physical world and, thereby, fostering both a deeper understanding of the natural and built environment and, ultimately, enhancing our ability to design and control these complex systems."

For an example of how the center's work on "urban sensing" can inform the interaction between society and the environment see http://urban.cens.ucla.edu/

Back to the top of the page.


WHO:Andrew Gelman, Professor in the Departments of Statistics and Political Science at Columbia University
WHAT: Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do
WHEN:10AM, Tuesday October 14th, 2008
WHERE:802 Kimmel Center for University Life, 60 Washington Square South
ABSTRACT:Andrew Gelman is a Professor in the Departments of Statistics and Political Science at Columbia University.  His new book, "Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do," is receiving tremendous critical praise. Gelman has recently been featured on several radio programs including WNYC's Leonard Lopate Show.
Professor Gelman recently appeared on the Leonard Lopate show; his talk will draw from his book on the same topic.

Back to the top of the page.