Gaussian processes Correlated Bayesian Additive Regression Trees

Xuetao Lu a,Robert E. McCulloch

arxiv(2023)

引用 0|浏览1
暂无评分
摘要
In recent years, Bayesian Additive Regression Trees (BART) has garnered increased attention, leading to the development of various extensions for diverse applications. However, there has been limited exploration of its utility in analyzing correlated data. This paper introduces a novel extension of BART, named Correlated BART (CBART). Unlike the original BART with independent errors, CBART is specifically designed to handle correlated (dependent) errors. Additionally, we propose the integration of CBART with Gaussian processes (GP) to create a new model termed GP-CBART. This innovative model combines the strengths of the Gaussian processes and CBART, making it particularly well-suited for analyzing time series or spatial data. In the GP-CBART framework, CBART captures the nonlinearity in the mean regression (covariates) function, while the Gaussian processes adeptly models the correlation structure within the response. Additionally, given the high flexibility of both CBART and GP models, their combination may lead to identification issues. We provide methods to address these challenges. To demonstrate the effectiveness of CBART and GP-CBART, we present corresponding simulated and real-world examples.
更多
查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要