Richard Emsley, Centre for Biostatistics.
Causal inference is concerned with the quantifying the relationship between a particular exposure (the ‘cause’) and an outcome (the ‘effect’). Implicitly or explicitly, causal inference is the primary aim of most empirical investigations, especially in medicine and behavioural science.
It can be summarised as explicitly defining the estimand of interest, formalising the assumptions required for traditional statistical models to estimate causal parameters and developing new statistical models and theory to estimate causal parameters. There has been a huge growth in publications relating to causal inference in the literature in the previous three decades.
In this talk we will:
- Give a brief recent history of causal inference
- Introduce the main concepts underpinning the dominant causal inference approach, known as potential outcomes or counterfactuals
- Give a general overview of the assumptions required for statistical models including structural equation modelling to produce causal effects
- Discuss a new set of causal estimands relating to mediation analysis
- Discuss a new class of models specifically for causal inference
- Highlight expertise and opportunities available at Manchester, including the first UK causal inference meeting
The talk is not overly technical, and focuses on the concepts rather than statistical theory so as to be accessible to everyone.
Download PDF slides of the presentation 'What is causal inference?'