Speaker：Assoc. Prof. Li-Pang Chen (Department of Statistics, NCCU)
Topic：Graphical estimation for multivariate error-prone variables with the availability of auxiliary information
Speaker：Assoc. Prof. Li-Pang Chen (Department of Statistics, NCCU)
Time：Oct 31 (Friday) , 2025, 10:40-11:30
Place： 4F-427, Assembly Building I
Abstract
Gene expression data in bioinformatics studies often involve multivariate or high-dimensional variables. A key research objective is to uncover the underlying network structure among gene expression variables, as this can help identify pathway-level disruptions associated with diseases and inform the development of targeted therapies. With the growing availability of auxiliary variables (referred to as covariates), it has become desirable to incorporate them to improve the detection of network structures among the primary variables (referred to as responses). The main challenge lies in accurately selecting informative covariates and recovering the network structure among the responses, particularly when employing linear or nonlinear models to characterize the relationships between multivariate responses and covariates. An additional complication arises from measurement errors in gene expression data, which may stem from limited measurement precision or human recording mistakes. In this talk, I will present several estimation strategies designed to address these complexities in both linear and nonlinear modeling settings. I will also discuss some theoretical properties of the proposed methods and present numerical studies to evaluate their performance.