Introduction to Causal Inference and Causal Diagrams (DAGs)

Although the ultimate research goal is to learn about causal effects, causality is rarely presented or discussed in statistical courses, except for the well-known mantra “association does not imply causation”. We learn that causal effects can only be estimated from randomized controlled experiments because in observational studies (i.e. studies that lack random allocation) alternative explanations for the relationship between “cause” and “effect” can’t be ruled out. However, some causal questions can only be answered from studies where randomization is not possible due to practical or ethical reasons.

This workshop will present an introduction on ways to enhance causal inference under observational study designs. We will discuss how to articulate the scientific question, including the definition of the causal parameter of interest, and will describe the set of conditions needed to estimate the causal effect of interest.

We will introduce a type of causal diagrams (DAGs), which are helpful tools to: i) communicate assumptions about our beliefs on how nature generates the data; ii) uncover sources of bias; and iii) guide data analysis.

Using causal DAGs in real and simulated data we will identify when one should or should not adjust for covariables, and which covariables one should adjust for.

Prerequisites

The workshop will assume basic knowledge of statistical methods such as regression models and basic knowledge of epidemiology.

Who should attend?

PhD students and researchers within the Faculty of Health.

Presented by

Liliana Orellana, PhD, Professor of Biostatistics

Contact

Please contact health-od@deakin.edu.au if you have any questions.