Colloquium Series: Paul Attewell, City University of New York

How Much Social Structure Is There?: The Challenge of Data Mining

Presented by Paul Attewell
Distinguished Professor of Sociology
City University of New York

Quantitative studies in the social sciences frequently report models that explain very small proportions of the variance in the dependent variable. This paucity of predictive power has elicited criticism by non-quantitative researchers and also bolsters perceptions that social structure, to the extent that it exists, is fleeting and situationally contingent. But how would sociologists’ views of social structuration change if quantitative models were much stronger? Data Mining methods can, in some cases, result in much higher predictive power than the conventional statistical approach. This talk reviews several aspects of data mining methods that often greatly improve model fit and, thereby shift our understanding of social structuration.

About the Speaker:
Paul Attewell is a Distinguished Professor of Sociology at the Graduate Center of the City University of New York, where he teaches courses on the sociology of education, on stratification and inequality, and on quantitative research methods. His current research focuses on processes affecting students from low-income backgrounds in non-elite colleges. His book Passing the Torch: Does Higher Education for the Disadvantaged Pay Off across the Generations? (Co-authored with David E. Lavin) won both the American Education Research Association’s Outstanding Book Award, and the Grawemeyer Prize in Education.

His methodological work, supported by an interdisciplinary grant from the National Science Foundation, concerns the application of machine-learning and data mining methods to social science and educational data.  One recent product of that project is a book co-authored by Professor Attewell and David Monaghan titled ‘Data Mining for the Social Sciences:  An Introduction’ (Univ. of California Press 2015).