A1196
Title: Tensor association test in Cox regression model
Authors: Chin-Chun Chen - National Cheng Kung University, Tainan (Taiwan) [presenting]
Pei-Fang Su - National Cheng Kung University (Taiwan)
Abstract: In survival analysis, representing multi-omics data as a tensor structure diversifies the types of covariates in the Cox regression model. Due to the complexity and high dimensionality of tensor covariates, a tensor rank decomposition technique is utilized for dimensionality reduction. A block relaxation algorithm, based on partial likelihood, is developed to iteratively estimate the effects of both clinical and tensor covariates. The innovation is a tensor association test, which facilitates statistical inference between survival outcomes and the two types of covariates. The proposed test enables the identification of genes within different platforms that demonstrate significant relationships with the survival outcomes. A comprehensive simulation study and real-case study on the Colorectal Adenocarcinoma TCGA PanCancer data are applied to demonstrate the method.
