What is a causal effect, and where can we find them? This talk will start from randomized controlled trials, why this (often) gives the quantity we want, and how the randomization aspect is important. This leads to a discussion about the difference between seeing versus doing, and the general problem of confounding, which can be seen in examples such as Simpson’s paradox. We will see that the problem of causal inference can be separated into two main questions - inferring the causal effect, and causal discovery. We look through some approaches for these two questions. We then briefly discuss how these ideas may help us think about predictive models.
Link to join virtually: https://cam-ac-uk.zoom.us/j/81322468305
This talk will not be recorded.
