Lectures and seminars SFOepi Seminar Series in Biostatistics: Kjetil Røysland
Title: Causal inference with counting processes: Graphical identifiability rules and efficient estimation.
Abstract: We consider continuous-time survival or more general event-history settings, where our aim is to graphically represent causal structures allowing us to answer queries such as when a causal parameter is identified from observational data. This causal parameter is formalised as the effect on an outcome event of a (possibly hypothetical) intervention on the intensity of a treatment process, i.e. a stochastic intervention. To establish identifiability, i.e. whether valid inference about the interventional situation can be drawn from typical observational (non-experimental) data, we propose novel graphical rules indicating whether the observed information is sufficient to obtain the desired causal effect by suitable re-weighting. This requires a different type of graph than in discrete time. We formally define causal semantics for the corresponding dynamic graphs that represent local independence models for multivariate counting processes. We propose a new method to calculate efficient influence functions for various causal parameters in these models, and then construct corresponding efficient estimators based on techniques from stochastic differential equations.
This is joint work with Vanessa Didelez and Pål C. Ryalen.