Skip to main content

Search NYU Steinhardt

Marc Scott

Professor of Applied Statistics

Applied Statistics, Social Science, and Humanities

212-992-9407

Dr. Scott's research involves the development of statistical models for longitudinal data. Using latent variable approaches, he develops new classes of covariance models for such data. He has used them to examine trends in wage inequality, with more recent applications including medical histories and psychological profiles. He also works on models for longitudinal sequence analysis. In the educational setting, such models examine the influence of the entire pathway (e.g., the timing of educational and employment spells and interruptions to these) on an outcome measure such as wages or degree completion. In work on low-wage labor markets, such models seek to find similar structure in the career histories and education of workers to identify more and less successful career ladders. More recent work involves multi-channel and model-based sequence analysis, sensitivity analysis for multilevel models and Bayesian computational methods.

Dr. Scott teaches Multi-Level Models: Growth Curves / Nested Data (APSTA-GE.2040/41/42), Practicum in Statistical Computing (APSTA-GE.2352), Supervised and Unsupervised Machine Learning (APSTA-GE.2011), and Spatial Statistics (APSTA-GE.2015). He serves as co-director with Lisa Stulberg of the Interdepartmental Research Studies (IDRS) program. In Fall 2008, Steinhardt launched a new applied statistics center, PRIISM (Center for Practice and Research at the Intersection of Information, Society, and Methodology), for which he and Jennifer Hill serve as co-directors. In 2014, Steinhardt launched a new Masters of Science in Applied Statistics for Social Research (MS-A3SR), for which he and Jennifer Hill serve as co-directors.

Selected Publications

  • Marc A. Scott, Ronli Diakow, Jennifer L. Hill, Joel A. Middleton. Potential for Bias Inflation with Grouped Data: A Comparison of Estimators and a Sensitivity Analysis Strategy. Observational Studies (2018).
  • Ying Lu, Sharon L. Weinberg, and Marc A. Scott. Repeated Measure Models and Longitudinal Models in Higher Education. In the Handbook on Measurement, Assessment, and Evaluation in Higher Education (2018). Routledge.
  • Marc A. Scott and Matthew Zeidenberg. Order or Chaos? Understanding Career Mobility Using Categorical Clustering and Information Theory. Longitudinal and Life Course Studies vol. 7 (2016), 320-346.
  • Joel A. Middleton, Marc A. Scott, Ronli Diakow, Jennifer L. Hill. Bias Amplification and Bias Unmasking. Political Analysis vol. 24 (2016), 307-323.
  • Ying Lu, Heng Peng, and Marc Scott. Penalized Fixed Effects Model for Clustered and Longitudinal Data. Proceedings of the 28th International Workshop on Statistical Modelling, (Jean-François Dupuy, Julie Josse, editors), vol. 1(2016), 171-176.
  • Ying Lu, Marc Scott, andPreeti Raghavan. Statistical Inference for Single-Case Design: Application to Post-stroke Rehabilitation. Proceedings of the 28th International Workshop on Statistical Modelling, (Jean-François Dupuy, Julie Josse, editors), vol. 1 (2016), 177-182.
  • Saul G. Alamilla, Marc A. Scott, and Diane L. Hughes. “The Relationship between Individual and Community-Level Sociocultural and Neighborhood Factors on the Mental Health of Ethnocultural Groups in Two Large U.S. Cities.” Journal of Community Psychology 44 (2016): 51–77.
  • Matthew Zeidenberg, Marc Scott and Clive Belfield. `What About the Non-Completers? The Labor Market Returns to Progress in Community College.' Economics of Education Review 49 (2015): 142–56.
  • Leoandra Onnie Rogers, Marc A. Scott, Niobe Way. “Racial and gender identity among Black adolescent males: An intersectionality perspective.” Child Development 86 (2015): 407-24.

Programs

Applied Statistics for Social Science Research

Learn advanced quantitative research techniques and apply them to critical policy issues across social, behavioral, and health sciences.

Courses

Applied Spatial Statistics

Spatial data arise when information collected on units that reside in different locations.Common examples include geology, criminology & epidemiology, where the goal may be to identify patterning or clusters (;hot spots') in the outcomes across the terrain being examined.In the social sciences, a similar set of questions & techniques are required, for example in studies of homelessness, poverty, environmental justice, & education.However, spatial data present a novel set of exploratory & modeling challenges, given the unique way in which outcomes are related (correlated) with each other through proximity.This course is an overview of the methods needed to analyze data for which it is suspected that the spatial component plays an important role.
Course #
APSTA-GE 2015
Units
2
Term
Fall
Faculty

Professors

Marc Scott ,
Department

Multi-Level Modeling Growth Curve

This is a course on models for multi-level growth curve data. These data arise in longitudinal designs, which are quite common to education and applied social, behavioral and policy science. Traditional methods, such as OLS regression, are not appropriate in this settings, as they fail to model the complex correlational structure that is induced by these designs. Proper inference requires that we include aspects of the design in the model itself. Moreover, these more sophisticated techniques allow the researcher to learn new and important characteristics of the social and behavioral processes under study. In this module, we will develop and fit a set of models for longitudinal designs (these are often called growth curve models). The course assignments will use state of the art statistical software to explore, fit and interpret the models.
Course #
APSTA-GE 2040
Units
2
Term
Fall
Faculty

Professors

Marc Scott ,
Department

Multi-Level Modeling: Nested Data/Longitudinal Data

This is a course on models for multi-level nested data. These data arise in nested designs, which are quite common to education and applied social, behavioral and policy science. Traditional methods, such as OSL regression, are not appropriate in this setting, as they fail to model the complex correlational structure that is induced by these designs. Proper inference requires that we include aspects of the design in the model itself. Moreover, these more sophisticated techniques allow the researcher to learn new and important characteristics of the social and behavioral processes under study. In this module, we will develop and fit a set of models for nested designs (these are sometimes called hierarchical linear models). The course assignments will use state of the art statistical software to explore, fit and interpret the models.
Course #
APSTA-GE 2042
Units
2
Term
Spring
Faculty

Professors

Marc Scott ,
Department

Practicum in Applied Statistics: Statistical Computing

This course will introduce the student to modern statistical programming & simulation using the language R.The core skills are oriented around first understanding variables, data structures, program flow (e.g., conditional execution, looping) & functional programming, then applying these skills to answer interesting statistical questions involving the comparison of groups, which is core to statistical practice.Most statistical analysis will be motivated via simulations, rather than mathematical theory.The course content (programming and data analysis) requires significant outside reading & programming.
Course #
APSTA-GE 2352
Units
1
Term
Fall
Faculty

Professors

Marc Scott ,
Department

Practicum in Multi-Level Models

This is a practicum course on models for multi-level growth curve data. This course is a natural sequel to E10.2040 Multi-Level Modeling Growth Curve. Building on the theory and examples developed in that course, students will participate in a guided, larger research project that employs multi-level growth curve models. Students will meet in groups with the instructor in a lab setting to fit, evaluate and describe these models. The final project for the course will consist of a 'results and discussion' section, journal article quality write-up.
Course #
APSTA-GE 2041
Units
1
Term
Fall
Faculty

Professors

Marc Scott ,
Department

Supervised and Unsupervised Machine Learning

Classification and clustering are important statistical techniques commonly applied in many social and behavioral science research problems. Both seek to understand social phenomena through the identification of naturally occurring homogeneous groupings within a population. Classification techniques are used to sort new observations into preexisting or know groupings while clustering techniques sort the population under study into groupings based on their observed characteristics. Both help to reveal hidden structure that may be used in further analysis. This course will compare and contrast these techniques, including many of their variations, with an emphasis on applications.
Course #
APSTA-GE 2011
Units
2
Term
January, Spring
Faculty

Professors

Marc Scott ,
Department