DART: Distance Assisted Recursive Testing
Xuechan Li, Anthony D. Sung, Jichun Xie; 24(169):1−41, 2023.
Abstract
In the field of modern data science, multiple testing is a widely used tool. In certain cases, the hypotheses are organized within a space where the distances between them reflect their co-null/co-alternative patterns. By properly incorporating the distance information into the testing process, the testing power can be significantly enhanced. This led to the development of a new multiple testing framework called Distance Assisted Recursive Testing (DART). DART is characterized by its joint utilization of artificial intelligence (AI) and statistical modeling. It consists of two stages. In the first stage, AI models are employed to construct an aggregation tree that captures the distance information. In the second stage, statistical models are used to embed the testing on the tree and control the false discovery rate. Both theoretical analysis and numerical experiments have demonstrated that DART produces valid, robust, and powerful results. To illustrate its application, we utilized DART in a clinical trial involving allogeneic stem cell transplantation to identify the gut microbiota that were impacted by post-transplant care.
[abs]