Bayesian Non Parametric Causal Inference in PyMC
probability
pymc
non parametric
causal inference
Bayesian Non Parametric Causal Inference
In this project I demonstrated how to fit and use BART propensity score models in the assessment of causal inference questions. We showed the conditions under which these methods work well and an example of where they break down due to overfit. We showed how they could be fixed using debiasing techniques and how appropriate analysis of this case could be achieved using non-parametric mediation analysis.
The project culminated in a publication to the official PyMC documentation that can be found online here or downloaded as notebook here
I also recorded a youtube modeling webinair on this topic with Alex Andorra
