A Gentle Introduction to Causal Inference
Online
In life, humans are typically very good at working out what causes something to happen, yet typically it is very hard for us to write down why we know this. This becomes even more complicated when data is involved. Therefore, in this course we will learn about the field of Causal Inference. This will start with looking at Pearl’s ladder of causality and understanding data concepts such as ‘confounders’ and ‘mediators’ before lightly touching upon the statistics needed to calculate the average causal effect (ACE) - how much one thing causes another to change. The course in the second week will then look at understanding the stable unit treatment value assumption (SUTVA) and how propensity scoring and the IPTW can help create more sophisticated estimation.
In each of the two sessions, the second half of session will be entirely practical with participants undertaking a range of tasks in either R or Python (knowledge of only one language is needed to undertake the course). For this the University’s Noteable service will be used to allow participants to quickly access the Python/R environments and get hands on practice without technological constraint. For this, participants will need access to their own laptop/computer.
Optional Additional Reading
If you are keen to understand more about this topic either before or after the course, then some recommended texts are below. For those intrigued more about the concept of causal inference (and without mathematical/statistical backgrounds) the Pearl text serves as a gentle introduction to the topic.
- The Book of Why (Pearl, 2018) [Introduction available online: http://bayes.cs.ucla.edu/WHY/]
- Causal Inference: What If (Hernán and Robins, 2023). [Available online: https://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/]
- Causal graph analysis of COVID-19 observational data in German districts reveals effects of determining factors on reported case numbers (Steiger, Mussgnug and Kroll, 2021). [https://doi.org/10.1371/journal.pone.0237277]
This is an intermediate-level course. No knowledge of the topic is required/expected, but it is expected that the learner has some knowledge of either R or Python (use of either RStudio/Jupyter Notebooks and basic packages like tidyverse/pandas).
This course will be taught by Chris Oldnall.
After taking part in this event, you may decide that you need some further help in applying what you have learnt to your research. If so, you can book a Data Surgery meeting with one of our training fellows.
More details about Data Surgeries.
Those who have registered to take part will receive an email with full details on how to get ready for this course.
If you’re new to this training event format, or to CDCS training events in general, read more on what to expect from CDCS training. Here you will also find details of our cancellation and no-show policy, which applies to this event.
Learning Outcomes:
- Demonstrate an understanding of the causal inference framework, including the concepts of Pearl’s ladder of causality, the average causal effect and the SUTVA.
- Be able to program approach data in R or Python and apply causal inference thinking to ensure proper estimation.
- Differentiate between different schools of thought in the causal inference field and appreciate the need for different methods in different scenarios.
You may be interested in these other training events:
- Digital Method of the Month: Machine Learning
- An Introduction to Machine Learning
- Null Hypothesis Testing in R
- Regression and Mixed Effects Modelling
