BNQD ow: A Bayesian non-parametric quasi-experimental design library using GP ow

Abstract

In this thesis we investigate the performance of a re-implementation of BNQD: a library for Bayesian Non-parametric Quasi-experimental Designs. BNQD functions as a Bayesian framework for quasi-experiments, and is capable of inferring causality under di erent assumptions than the ones required by a randomised controlled trial. We will compare the performance of this new implementation to the previous one, as well as examine the e ectiveness of the features that were previously unavailable.