Multilevel Modeling of Social Problems: A Causal Perspective

Robert B. Smith

2011年2月 · Springer Science & Business Media

关于此电子书

Contemporary societal problems are complex, intractable, and costly. Aiming to ameliorate them, social scientists formulate policies and programs, and conduct research testing the efficacy of the interventions. All too often the results are disappointing; partly because the theories guiding these studies are inappropriate, the study designs are flawed, and the empirical databases covering their research questions are sparse. This book confronts these problems of research by following this process: analyze the roots of the social problem both theoretically and empirically; formulate a study design that captures the nuances of the problem; gather appropriate empirical data operationalizing the study design; model these data using multilevel statistical methods to uncover potential causes and any biases to their implied effects; use the results by refining theory and by formulating evidence-based policy recommendations for implementation and testing.

Applying this process, the chapters focus on these social problems: political extremism; global human development; violence against religious minorities; computerization of work; reform of urban schools; and the utilization and costs of health care. Because these chapters exemplify the usefulness of multilevel modeling for the quantification of effects and causal inference, they can serve as vivid exemplars for the teaching of students. This use of examples reverses the usual procedure for introducing statistical methods. Rather than beginning with a new statistical model bearing on statistical theory and searching for illustrative data, each core chapter begins with a pressing social problem. The specific problem motivates theoretical analysis, gathering of relevant data, and application of appropriate statistical procedures. Readers can use the provided data sets and syntaxes to replicate, critique, and advance the analyses, thereby developing their ability to produce future applications of multilevel modeling.

The chapters address the multilevel data structures of these social problems by grouping observations on the micro units (level-1) by more macro-units (level-2) (e.g., school children are grouped by their classroom), and by conducting multilevel statistical modeling in contextual, longitudinal, and meta-analyses. Each core chapter applies a qualitative typology to nest the variance between the macro units, thereby crafting a "mixed-methods" approach that combines qualitative attributes with quantitative measures

作者简介

Robert B. Smith (Ph.D. Columbia University, 1971) taught political sociology, research methods, and theory development at the University of California, Santa Barbara. His research there focused on the social consequences of war, generalizations of path analysis, and computer simulations of social processes. Since then, he has worked extensively in applied research. His publications include articles on political and social processes, and on multilevel models bearing on human development. He was the primary editor of the three volumes of the A Handbook of Social Science Methods, which link qualitative and quantitative methods, and he is the author of Cumulative Social Inquiry: Transforming Novelty into Innovation. His recent research at the Cambridge-MIT Institute assesses student exchange programs and pedagogical experiments. As senior statistician he worked on software for randomized trials, exact statistics, and Bayesian simulations at Cytel Inc., and is an advisory editor of Quality & Quantity. He was a Fulbright lecturer in structural sociology at Ghent University, Belgium, and he has served as president of the Boston chapter of the American Statistical Association. He resides in Cambridge, Massachusetts where he directs his social structural research.