Dr. Chixiang Chen

Title: Robust Causal Machine Learner in Mean Estimation and information integration from Auxiliary Data

Abstract: Modern machine learning algorithms are compelling in prediction problems. However, regarding the black-box feature, the performance of machine learning algorithms is hard to be statistically evaluated and could substantially vary across databases and underlying setups. As a result, locating the most appropriate algorithm will become notoriously challenging when multiple algorithms, such as penalized regression, random forest, gradient boosting, etc., are considered and evaluated, especially in the context of causal inference. In this talk, the first goal is to propose a robust causal machine learner in the context of mean estimation. The proposed learner enables valid statistical inference and has robustness property: multiple machine learning algorithms are allowed, and it has shown to be robust as long as one of candidate algorithms works well. The second goal in this talk is to build a data-integration scheme of synthesizing extra information from auxiliary data into the casual machine learner, which can substantially boost the estimation efficiency. Extensive numerical studies demonstrate the superior of our method over competing methods, in terms of smaller estimation bias and variability. Moreover, the validity of the proposed method is assessed in a real study of brain-aging by using UK Biobank data.