Abstract

Speaker: Dr Yordan Raykov, School of Mathematical Sciences, University of Nottingham, UK (Visiting in person)

Title: Gentle introduction to causal inference and estimation of causal effects in the wild

Abstract: This hands on tutorial walks participants from first principles to state of the art practice in causal inference. We begin by showing how the back door and front door criteria; the causal direction of an association and how we can determine whether an effect is identifiable. We then contrast classic parametric estimators with modern non parametric, machine learning based approaches, explaining when prediction accuracy matters, when it does not, and why. Next, we tackle unmeasured confounding by presenting instrumental variable and negative control strategies, as well as how classical latent variable models can be used in many situations. A rapid tour of constraint based, score based, and functional causal model algorithms, together with the Python packages that implement them, illustrates how to move from data to a working causal graph. Finally, we survey front line challenges such as longitudinal data with time varying confounding, irregular sampling schedules, and high dimensional outcomes like feature vectors, brain connectome matrices, or wearable device curves. By the end of the session hopefully you will know when a causal question is answerable, how to estimate its effect using the right tool, and where to find some of the software that lets you put these ideas into practice.

Bio: Dr Yordan Raykov is an Assistant Professor in Data Science at the University of Nottingham. His research develops novel statistical machine learning algorithms for clustering, dimensionality reduction, feature sharing, and sequence modeling, with applications in digital sensor monitoring and precision medicine. Before joining Nottingham, he worked in both academia and industry, including with ARM Cambridge, Radboudumc, UCB Pharma, Johns Hopkins, and the Michael J. Fox Foundation.

Venue: MRC CBU West Wing Seminar Room and Zoom https://us02web.zoom.us/j/82385113580?pwd=RmxIUmphQW9Ud1JBby9nTDQzR0NRdz09 (Meeting ID: 823 8511 3580; Passcode: 299077)