Accounting for shared covariates in semi-parametric Bayesian additive regression trees

Abstract

We propose a new semi-parametric model based on Bayesian Additive Regression Trees (BART). In our approach, the response variable is approximated by a linear predictor and a BART model, where the first component is responsible for estimating the main effects and BART accounts for the non-specified interactions and non-linearities. The novelty in our approach lies in the way we change tree generation moves in BART to deal with confounding between the parametric and non-parametric components when they have covariates in common. Through synthetic and real-world examples, we demonstrate that the performance of the new semi-parametric BART is competitive when compared to regression models and other tree-based methods. The implementation of the proposed method is available at https://github.com/ebprado/SP-BART.

Learning counterfactual outcomes of MOOC multiple learning behaviors

“…This method can handle a large number of predictor variables, is applicable to both continuous-type treatments and missing data, and outperforms traditional propensity score matching in estimating the average causal effect. Prado proposed a new semiparametric estimation method based on BART, where the estimation of the main effects is performed by a linear predictor, and BART is improved to handle the confounding between parameters and nonparameters, combining these two components to approximate the potential outcome [19]. The second is the application of random forest in causal inference.…”

Bayesian Cart Models for Insurance Claims Frequency

“…Since BCART models and their ensemble version -the Bayesian Additive Regression Trees (BART) models -generally outperform other machine learning models, they have been extensively studied in the literature; see, e.g., Linero (2017); Chipman et al (2010); Prado et al (2021); Murray (2021); Hill et al (2020) and references therein. In particular, their excellent empirical performance has also motivated works on their theoretical foundations; see Linero and Yang (2018); Rocková et al (2020).…”