Graphical Causal Models for Social Scientists
A 2-Day Seminar Taught by Felix Elwert, Ph.D.
Graphical causal models empower applied social and biomedical scientists to bridge the gap between statistics and causality—between data and theory. The gap between statistics and causality animates many fierce debates in the social sciences: For example, social scientists know that married people live longer; but does marriage actually prolong life? They know that graduates of elite colleges earn more; but do they earn more because they graduated from elite colleges? They know that mothers earn less than childless women, but does that descriptive difference really capture the full cost of motherhood?
Causal questions are questions about making a difference, about producing change. Answering causal questions, however, is famously difficult. Regrettably, statistical associations are blind to causality. Whether or not a regression can be interpreted as a causal effect always hinges on the analyst’s theory of data generation—how can we reason backward from the observed data to understand how the world works?
This seminar offers a far-ranging applied introduction to graphical causal models (directed acyclic graphs, DAGs) for social and biomedical scientists. Over the past few years, DAGs have become a popular tool for understanding and resolving causal problems. The chief advantage of DAGs is that they are “algebra-free.” DAGs do heavy mathematical lifting without looking like math. Instead, DAGs plainly encode the ordinary cause-and-effect statements that drive everyday scientific conversations. DAGs’ accessibility makes them popular with applied researchers, and their rigor makes them attractive to methodologists as well.
The three primary uses of DAGs are (1) notating the analyst’s theory, (2) deriving its testable implications, and (3) determining whether and which causal effects in the theory can actually be “identified” from the data. Among other things, DAGs clarify the conditions under which regression coefficients can be interpreted as causal effects. DAGs also reveal the logic and challenges of other widely used social science techniques, such as matching, instrumental variables, and mediation analysis.
This seminar focuses on building transferable intuition and skills for applied causal inference. We start by introducing the essential elements for causal reasoning with graphical causal models and then use DAGs to discuss a range of important challenges in observational data analysis. Topics include: conditions for the identification of causal effects; the difference between confounding bias, over-control bias, and selection bias; what variables to control for; what variables not to control for; specifying regression models; instrumental variables analysis; and mediation analysis.
Who should attend?
If you want to understand under what circumstances social scientists can draw causal inferences from non-experimental data, this course is for you. Participants should have a good working knowledge of multiple linear regression. Some basic knowledge of probability theory will be helpful but is not essential. You do not need to know matrix algebra or calculus.
Since this seminar aims to strengthen your ability to think through causal problems, we will do numerous pencil-and-paper exercises. You will learn all necessary technical tools in this seminar. There is no software component.
LOCAtions, Format, And Materials
The class will meet from 9 am to 4 pm each day with a 1-hour lunch break at Temple University Center City, 1515 Market Street, Philadelphia, PA 19103.
Participants receive a bound manual containing detailed lecture notes (with equations and graphics), examples of computer printout, and many other useful features. This book frees participants from the distracting task of note taking.
Registration and lodging
The fee of $995.00 includes all seminar materials. The early registration fee of $895 is available until July 9.
If you cancel your registration at least two weeks before the course is scheduled to begin, you are entitled to a full refund (minus a processing fee of $50).
Lodging Reservation Instructions
A block of guest rooms has been reserved at the Club Quarters Hotel, 1628 Chestnut Street, Philadelphia, PA at a special rate of $137 per night. This location is about a 5 minute walk to the seminar location. In order to make reservations, call 203-905-2100 during business hours and identify yourself by using group code STA808 or click here. For guaranteed rate and availability, you must reserve your room no later than Monday, July 9, 2018.
If you make reservations after the cut-off date, ask for the Statistical Horizons room rate (do not use the code) and they will try to accommodate your request.
a. The potential-outcomes (counterfactual) model of causality
b. Why causal inference requires causal assumptions (i.e. social theory)
2. Directed Acyclic Graphs (DAGs)
a. Graphs as a visual representation of social theory
b. Displaying causal assumptions in a graph
3. From causation to association
a. Why some variables are associated and others aren’t (“d-separation”)
b. Causation, confounding, and selection bias
4. From association to causation
a. What associations (e.g. regression coefficients) capture causal effects
b. The backdoor criterion: a powerful rule for deciding identification
c. Fast—and safe—shortcuts to identification
5. Selection Bias with lots of examples
a. Selecting on the outcome
b. Selecting on mediators
c. Selection bias from controlling for baseline variables
d. Why selection and confounding aren’t the same thing
6. Graphs for regression
a. How regression and matching try to identify causal effects
b. Which regression coefficients have a causal interpretation?
c. Wright’s method of path coefficients—quantifying biases in linear models
7. Instrumental variables analysis
a. Using graphs to understand IV
b. Graphical intuition for understanding exclusion violations in practice
8. Causal mediation analysis
a. Controlled vs natural effects
b. The fundamental problem of mediation is selection on the mediator
c. What variables to control for in mediation analysis.
“I have spent years reading literature on this topic and Dr. Elwert virtually provided a complete illumination of Causal Inference with Directed Graphs in just two days. Save yourself time, struggles, and confusion by taking this wonderful class!”
Joseph Staats, University of Minnesota Duluth
“Felix is an excellent instructor of this course and going through his training has helped me to understand the important basics of DAGs and graphical analysis as well as their importance in causal inference. I would highly recommend this course to anyone involved in causal inference in whatever area of specialty because, believe you me, this instructor is good!”
Edmore Chamappiwa, University of Manchester, UK
“Every scientist engaged in biomedical research should take this course. Understanding causal inference is critical for conducting research that can impact the quality of life of patients, advance the science of medicine, and promote the health and welfare of all.”
Debbie Barrington, Georgetown University
“As an epidemiologist who was not taught DAGs in their training, and who has attempted to read articles explaining the rules of DAGS, this class was great. Left me feeling more empowered in my discipline.”
“My understanding of research design changed after this course. This is one of the most powerful courses in methods I have ever encountered.”
Anibal Perez-Linan, University of Pittsburgh
“This course helps reinforce epidemiologic methods through DAGs. It was extremely valuable, fun and worth participation. It gave me a great mental tune-up.”
Ximena Vergara, Electric Power Research Institute/UCLA
“Causal Inference with Directed Graphs provides a better alternative in making causal inferences from regression analysis. The instructor is good and experienced at the subject and you can be sure of a better understanding at the end of the workshop.”
Godfred Boateng, Cornell University
“The course opened my eyes to some causal problems linked to ordinary least squares regression analysis. It was also impressive that Professor Elwert knew our names that fast.”
Henk Von Eije, University of Groningen
“Felix is an amazing teacher. Explains the basics (and advanced) stuff with such ease. I have been reading a lot to understand causality for a long time, but this course made it look simple and cool. I really want to take my learning forward by applying DAG to test my results from previous analysis.”
Vivek Kumar Sundriyal, Lund University