BARP: Improving Mister P Using Bayesian Additive Regression Trees
BARP: Improving Mister P Using Bayesian Additive Regression Trees
复制标题
DOI：
10.1017/s0003055419000480
复制
发表时间：
2019-11-01
影响因子：
6.8
通讯作者：
Bisbee, James
中科院分区：
文献类型：
--
作者：
Bisbee, James
多级回归和后分层 (MRP) 是将全国代表性调查的意见数据外推到较小地理单位的当前黄金标准。然而，非参数正则化方法的创新可以进一步提高研究人员将意见数据外推到感兴趣的地理单位的能力。我测试了一组正则化算法，发现通过更灵活的正则化方法可以对多级模型进行实质性改进。我提出了 MRP 的修改版本，它用称为贝叶斯加性回归树（BART，或与后分层结合时的 BARP）的非参数方法取代了多级模型。我在多个数据环境中比较了这两种方法，证明了应用更强大的正则化方法将意见数据推断到目标地理单位的好处。我提供了一个实现 BARP 方法的 R 包。
Multilevel regression and post-stratification (MRP) is the current gold standard for extrapolating opinion data from nationally representative surveys to smaller geographic units. However, innovations in nonparametric regularization methods can further improve the researcher's ability to extrapolate opinion data to a geographic unit of interest. I test an ensemble of regularization algorithms and find that there is room for substantial improvement on the multilevel model via more flexible methods of regularization. I propose amodified version ofMRPthat replaces the multilevel model with a nonparametric approach called Bayesian additive regression trees (BART or, when combined with post-stratification, BARP). I compare both methods across a number of data contexts, demonstrating the benefits of applying more powerful regularization methods to extrapolate opinion data to target geographical units. I provide an R package that implements the BARP method.