GESIS Training Courses
user_jsdisabled
Search

Scientific Coordination

Dr.
André Ernst
Tel: +49 221 4703736

Administrative Coordination

Janina Götsche

Directed Acyclic Graphs for Causal Inference

About
Location:
Cologne / Unter Sachsenhausen 6-8
 
General Topics:
Course Level:
Format:
Software used:
Duration:
Language:
Fees:
Students: 300 €
Academics: 450 €
Commercial: 900 €
 
Keywords
Additional links
Lecturer(s): William Lowe

About the lecturer - William Lowe

Course description

The course offers a comprehensive overview of causal identification and inference based around Directed Acyclic Graphs (DAGs) or causal graphs. DAGs provide a unified and intuitive graph-theoretic framework for thinking about causal inference, and are particularly well suited for social science problems where only partial information about mechanisms and functional forms is available, where experimentation is difficult or impossible, and where the research focus is explanation rather than effect estimation. DAGs can also illuminate common causal structures and assumptions behind many familiar but apparently disparate statistical methods and research designs, from regression to matching, missing data, sample selection bias, multilevel modeling, and measurement error, as well as illuminating normative topics such as race and gender bias and algorithmic fairness.

While the course is focused on theory and many exercises will require no more than a problem statement and a whiteboard, we will also make use of software for drawing DAGs and deriving observational consequences from them, and use empirical simulation and estimation to see the practical consequences of decisions made about DAG structures in a wide range of substantive problems. Teaching will be in person and take the form of lectures, short exercises, and - most importantly - discussion.


Target group

Participants will find the course useful if:
  • they work on problems that require causal inference from observational data
  • need to evaluate study designs for causal inference


  • Learning objectives

    By the end of the course participants will:
  • be able to translate research problems into assertions about causal structures and estimands
  • understand the motivating causal assumptions behind a wide range of existing statistical tools, and the ways in which they may fail to hold
  •  
    Organisational structure of the course
  • Work on assignments, individual or group work on participants' projects, group discussions, and hands-on exercises.
  • The lecturer will be available for individual and group discussion of participants' projects and to support work on assignments in the lab.


  • Prerequisites

  • A conceptual understanding of experimentation, randomization, and statistical control
  • Familiarity with and ability to manipulate statistical expectation, independence, and conditional independence, and joint, marginal, and conditional distributions, up to Bayes theorem. We will make almost no assumptions about functional forms, so familiarity with the algebraic details of particular probability distributions is not required.
  • Experience with fitting and interpreting multivariate regression models, including interactions in the R language.
  • A willingness to revisit and rethink the content of previous statistics courses(!)
  •  
    Software and hardware requirements
    Please bring a laptop or other device with R and RStudio installed
    However, a detailed instruction will be sent beforehand via e-mail, and there will be points of contact available for troubleshooting


    Schedule

    Recommended readings