Abstract:
Bayesian additive regression tree (BART) models are becoming increasingly popular
in literature thank to their high precision and adaptability to many different data
settings. From the basic model, many improvements have been proposed, such as
BART models that adapt to sparsity in the data and smoothness in the underlying
true function. Others have attempted to provide models that intercept causation effects
and others to provide some background theory. In this thesis I propose an application
to economics, through BART estimation of the five factor asset pricing model by Fama
& French. The aim is to give evidence in favor or against the correct specification of
the model.
