Applied Causal Inference with Directed Acyclic Graphs
Course

Applied Causal Inference with Directed Acyclic Graphs

Starts Mar 27, 2025

$375 Enroll

Full course description

SHORT COURSE DESCRIPTION

This two-day course provides an applied introduction to directed acyclic graphs (DAGs) for causal inference. The primary goal is to teach the effective use of causal DAGs in applied research. Course participants learn to (i) draw valid causal graphs, (ii) determine the most promising causal identification strategy, (iii) choose valid sets of control variables, (iv) estimate the average causal effect, (v) assess the direction and extent of any remaining confounding bias, and (vi) identify endogenous sample selection bias and other collider bias issues that threaten causal inference.

While the focus is on applying DAGs to causal evaluation questions, the basic theory of DAGs and their underlying structural causal models is introduced alongside practical applications as needed. Hypothetical and real-world examples illustrate how to draw and analyze DAGs, identify average causal effects, and estimate them from actual data. Estimation techniques covered include regression, inverse-probability-of-treatment weighting, and doubly robust estimators. The application of DAGs for causal mediation analysis and causal decomposition analysis in disparity research is briefly addressed. 

The course covers DAG terminology and key concepts including colliders & collider bias, blocking paths, d-separation, the do-operator, and the adjustment/backdoor criterion. The usefulness of graphs is demonstrated mostly for observational studies that rely on matching/propensity score adjustments, but alternative identification strategies (instrumental variables, difference-in-differences/gain score) as well as non-compliance and attrition issues in RCTs are also discussed graphically. The dagitty online app and corresponding R package (https://www.dagitty.net/) is used to draw and analyze DAGs. Applied data analyses are demonstrated in R.

 

DATES AND TIMES

March 27-28, 2025 (Thurs-Fri)

10am-5pm Eastern Daylight Time (UTC-4)

The instructor will determine timing of breaks.

 

COURSE FEES

Professional: $375

Full-time student*: $195

 

*Full-time students need to submit student status proof at https://go.umd.edu/CILVR-Student-24 to request a discount code prior to registration.

 

*Course fee will be waived for HDQM Department faculty and degree-seeking students, although the UMD IT department will charge you a tech fee to register ($10). HDQM department registrants can request the discount code by submitting the following formhttps://go.umd.edu/CILVR-HDQM-24.

 

 

HOW TO REGISTER

 

Prior to registration, participants not affiliated with UMD need to get a valid UMD associate account in order to register for the short course and access the course content. Participants can visit https://identity.umd.edu/id/associate/registration to create an UMD associate account. For more details about the UMD associate account, please click here.

 

For UMD affiliated participants, you may register using your existing UMD directory ID. 

 

To request the promotional code prior to registration: 

- Full-time students can submit the student status proof at https://go.umd.edu/CILVR-Student-24 to request a student discount code prior to registration. Note that it may take 2-3 business days for your request to be processed.

 

- HDQM department registrants can request the HDQM discount code by submitting the following formhttps://go.umd.edu/CILVR-HDQM-24Note that it may take 2-3 business days for your request to be processed.

 

TARGET AUDIENCE

Graduate students, faculty, research professionals who are interested in the theory and application of causal inference and directed acyclic graphs. 

 

REQUISITE KNOWLEDGE

  • Intermediate proficiency in statistical programming language (e.g., R)
  • Intermediate proficiency in inferential statistics

 

 

SOFTWARE

 

The dagitty online app and corresponding R package (https://www.dagitty.net/) is used to draw and analyze DAGs. Applied data analyses are demonstrated in R.

 

 

LOCATION AND PLATFORM

 

  • The course materials and meeting links will be posted on the course page through UMD Open Learning (https://umd.catalog.instructure.com/)
  • This workshop will be delivered entirely online via the video conferencing software Zoom (https://zoom.us/).
  • Within a limited time, the video recordings of the short course will be accessible for both synchronous and asynchronous participants on the course page.

 

IMPORTANT COURSE DETAILS

Platform: Participants need to have a valid UMD associate ID in order to register for the short course and access the course content. Participants can visit https://identity.umd.edu/id/associate/registration to create an UMD associate account. For more details please click here.

 

Format: Participants will receive a personalized login code to use on their own computer to access a reliable live-stream of the short course over Zoom, showing the instructor as well as the handouts.

 

Materials: Participants will receive electronic copies of the short course materials, as well as any other relevant materials or information.

 

Timing/access: Participants may choose to watch the stream synchronously, or may elect to watch a recording of the short course asynchronously, or both. Recordings will be available to participants for six months following the end of the short course. This is especially useful for on-line participants in different time zones who may choose to watch at some later time than (but within six months of) the actual short course time. (Asynchronous participation does not include real-time chat with other on-line participants, although a visual record of prior chats will be viewable).

 

Technical support: Participants are responsible for installing the conferencing software Zoom on their own electronic devices and for obtaining a Zoom account that allows the participant to join Zoom meetings and webinars hosted by external organizations. Participants are assumed to be able to secure a reliable computer, internet browser, and Wi-Fi connection. Challenges at the user end must be resolved by the user. Fortunately, because the short course is recorded, users experiencing technical challenges can still “catch up” by watching the recordings to which they have access.

 

Content support: During the lecture, real-time content support for on-line participants is mostly limited to real-time chat with the on-line (Zoom) participant community and any quantitative methodology doctoral students who might also be participating. Participants may have direct interactions with the instructor in some format during the practice sessions. On-line participants may e-mail the instructor for further content support that cannot be addressed in real-time.

 

THE INSTRUCTOR

Dr. Peter M Steiner is a Professor in the Quantitative Methodology: Measurement and Statistics (QMMS) program in the Department of Human Development and Quantitative Methodology at the University of Maryland. Prior to joining the QMMS faculty in fall of 2019, he was a faculty member of the Department of Educational Psychology at the University of Wisconsin-Madison (2010-2019), a research associate at the Institute for Policy Research at Northwestern University (2007-2010), and a researcher and Assistant Professor at the Institute for Advanced Studies in Vienna, Austria (1997-2007). Peter M Steiner received a master’s and doctorate degree in Statistics from the University of Vienna and a master’s degree in Economics from the Vienna University of Economics and Business Administration. His research on causal inference, replication, and factorial surveys has appeared in such journals as Psychological Methods, Multivariate Behavioral Research, Journal of Educational and Behavioral Statistics, Evaluation Review, Sociological Methods & Research, Journal of Causal Inference, or the Journal of the American Statistical Association. In 2019, he received the Causality in Statistics Education Award of the American Statistical Association. Dr. Peter Steiner can be reached at psteiner@umd.edu.

 

REFUND POLICY

Full refund if cancellation occurs at least 10 business days prior to the workshop date; 50% refund if within 10 days of the first day of the course.

 

CONTACT

For any further questions, please contact cilvr-shortcourses@umd.edu.

To request a copy of the payment receipt, please contact the OES office at oes-finance@umd.edu.

 

CILVR Short Course Series

 

Center for Integrated Latent Variable Research (CILVR) at the University of Maryland (UMD)

CILVR is a center whose goal is to serve as a national and international focal point for innovative collaboration, state-of-the-art training, and scholarly dissemination as they relate to the full spectrum of latent variable statistical methods. CILVR is housed within the Measurement, Statistics and Evaluation (EDMS) program in the Department of Human Development and Quantitative Methodology at the University of Maryland. EDMS faculty are recognized scholars in various facets of latent variable statistical models, whether it be item response theory, latent class analysis, mixture models, latent growth models, or structural equation modeling.