Causal Inference and Causal Mediation Analysis 《因果推论和因果机制分析》
课时：36学时 招生人数：80人 工作语言：英语
主讲教师：Guanglei Hong教授和Stephen Raudenbush教授
Guanglei Hong is Associate Professor with tenure in the Comparative Human Development Department and the Committee on Education at the University of Chicago. She is an active member of the quantitative methodology community at the University. Dr. Hong develops and applies causal inference theories and methods for evaluating educational and social policies and programs in multi-level, longitudinal settings. Her work is currently focused on developing concepts and methods for analyzing causal mediation mechanisms in multisite randomized trials. She has received research funding from the National Science Foundation, the Institute of Education Sciences of the U.S. Department of Education, the William T. Grant Foundation, the Spencer Foundation, and the Social Sciences and Humanities Research Council of Canada.
Her research monograph “Causality in a social world: Moderation, mediation, and spill-over” was published by John Wiley & Sons in July 2015. The book clarifies for applied researchers the theoretical concepts of moderated effects, mediated effects, and spill-over effects. It systematically introduces innovative statistical strategies for investigating these causal effects and aims to make them readily accessible to a broad audience. A major emphasis is placed on explicating and evaluating, in the context of real applications drawn from social sciences, education, and health research, the assumptions required for relating causal parameters of interest to empirical data given a specific research design. The book is accompanied by data examples and statistical programs for the new causal inference methods. Her other publications have appeared in statistics, education, and psychology journals including the Journal of the American Statistical Association, the Journal of Educational and Behavioral Statistics, Educational Evaluation and Policy Analysis, the Journal of Research on Educational Effectiveness, Psychological Methods, and Developmental Psychology, among others. She was Guest Editor for the Journal of Research on Educational Effectiveness special issue on the statistical approaches to studying mediator effects in education research published in 2012.
Stephen Raudenbush is the Lewis-Sebring Distinguished Service Professor in the Department of Sociology, the College and the Harris School of Public Policy Studies and Chairman of the Committee on Education at the University of Chicago. He is a member of the National Academy of Sciences and the American Academy of Arts and Sciences. He received the American Educational Research Association award for Distinguished Contributions to Educational Research.
He is interested in statistical models for child and youth development within social settings such as classrooms, schools, and neighborhoods. He is best known for his work developing hierarchical linear models, with broad applications in the design and analysis of longitudinal and multilevel research. He is currently studying the development of literacy and math skills in early childhood with implications for instruction; and methods for assessing school and classroom quality.
网址：https://qrm.uchicago.edu/directory/guanglei-hong and https://sociology.uchicago.edu/directory/stephen-raudenbush
The goal of this two-week course is to equip students with basic knowledge and analytic skills in causal inference and causal mediation analysis. During the first week, the course will introduce the potential outcomes framework, clarify the assumptions required for identifying causal relationships in experimental and observational studies, and discuss the potential pitfalls of using ANCOVA or multiple regression to estimate a causal effect. The course will then focus on a number of propensity score based methods including matching, stratification, and weighting. Several econometric methods for causal inference will receive a brief review before the end of the first week. The second week will start with the topic about identifying the average treatment effect in a multisite randomized trial, revealing that treatment randomization at each site does not guarantee an unbiased analytic result even when there is perfect compliance. Further challenges and solutions will be discussed in the context of noncompliance. The course will then introduce causal mediation analysis in single-site and multisite studies, contrasting several alternative analytic strategies, and focusing on a new weighting-based approach to mediation analysis. Last but not least, the course will conclude with a discussion of sensitivity analysis for assessing the potential impact of hidden bias.
(1) Does Schooling Increase or Reduce Social Inequality?
(2) Some New Tools for Causal Inference in Multilevel Research
General Designs and Sensitivity Analysis for Causal Inference 《因果推论方法的研究设计和敏感性分析》
课时：36学时 招生人数：80人 工作语言：英语
Kenneth Frank received his Ph.D. in measurement, evaluation and statistical analysis from the School of Education at the University of Chicago in 1993. He is currently a professor in Counseling, Educational Psychology and Special Education as well as in Fisheries and Wildlife and adjunct in Sociology at Michigan State University. His substantive interests include the study of schools as social organizations and the social embeddedness of natural resource use. His substantive areas are linked to several methodological interests: social network analysis, causal inference and multi-level models. His publications include quantitative methods for representing relations among actors in a social network, robustness indices for inferences, and the effects of social capital in schools and other social contexts. He teaches general introductory courses in research methods and quantitative methods as well as advanced courses in multivariate analysis and seminars in social network analysis and causal inference. Ken’s current projects include a study of the effects of the Michigan Merit Curriculum on educational outcomes and how knowledge about climate change diffuses to policy-makers and educators.
There is currently great debate regarding the basis for causal inferences across the social sciences. Can we make causal inferences only from experiments? What about ethical or logistical limitations, or concerns that the experimental paradigm is artificial because of the necessity for extreme control over conditions? On the other hand, though observational studies are applied to natural conditions, can we rely on statistical control to make causal inferences? What about unmeasured, or unrecognized confounding factors? At what point does a statistical inference sustain a causal inference? Answers to these questions are more than merely academic and philosophical. For example they have immediate implications for policy-making regarding the implementation of innovations.
To address questions such as the above this course will explore causal inference from the perspectives of statistics and philosophy of science. We will begin with a comparison of causal inferences in the social sciences with those of the experimental sciences. Drawing on eclectic readings (Manski, Heckman, Rubin, Holland, Pearl, Shadish, Cook, Campbell Sobel, Dawid), we will use concepts such as the counterfactual, homogeneity of units and internal and external validity to describe causal inference. Furthermore, we will discuss statistical techniques such as propensity score matching and instrumental variables that might be used to improve the likelihood of valid inferences. Finally, we will use recent work to quantify how robust inferences are to potential threats the validity.
In the first half the course I will present methods including regression, propensity score matching, instrumental variables, regression discontinuity, random versus fixed effects, and sensitivity analysis. In the second half of the course we will turn to intensive projects or readings. Students will be required to present and provide an intensive critique of an article featuring complex issues of causal inference. For the final project students will need to develop work toward a publishable paper, either in the form of new method, or existing method applied to data of the students’ interest. Students may work in groups, with indications for who had primary responsibility for what work.
电话：010-6276-6210 传真：010-6276-7623 电邮：firstname.lastname@example.org