Parameterizing and Simulating from Causal Models


Event details

Date 13.02.2023 14:3015:30  
Speaker Robin EVANS, University of Oxford
Location Online
Category Conferences - Seminars
Event Language English
Seminar in Mathematics
: Many statistical problems in causal inference involve a probability distribution other than the one from which data are actually observed; as an additional complication, the object of interest is often a marginal quantity of this other probability distribution. This creates many practical complications for statistical inference, even where the problem is non-parametrically identified. In particular, it is difficult to perform likelihood-based inference, or even to simulate from the model in a general way. 

We introduce the frugal parameterization, which places the causal effect of interest at its centre, and then builds the rest of the model around it. We do this in a way that provides a recipe for constructing a regular, non-redundant parameterization using causal quantities of interest. In the case of discrete variables we can use odds ratios to complete the parameterization, while in the continuous case copulas are the natural choice. 

Our methods allow us to construct and simulate from models with parametrically specified

Practical information

  • Informed public
  • Free
  • This event is internal


  • institute of Mathematics


  • Prof. Maryna Viazovska, Director

Event broadcasted in