A software package supporting do-calculus and other causal inference tools.
This project began under the supervision of Dr. Neufeld in May 2020 while I was a Summer Research Assistant at the University of Saskatchewan. This role ran from May 2020 to August 2020, though I have continued as a Student Research Assistant since. I have enjoyed this opportunity to work with Dr. Neufeld on this project (and a few other things!) and learned much in the process.
Scope / Scale
This project implements the entire do-calculus of Judea Pearl et. al, as well as the identification algorithm presented in Shpitser and Pearl, 2006. This work takes a graphical approach to enable the measurement of interventional distributions from observational data. This category of work is essentially a heavily-modified Bayesian network, and so this project is also a fully-functional Bayesian net, enabling standard probability queries as well.