Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 83 - Frontiers in Analysis of Microbiome Data: From Methods to Applications
Type: Invited
Date/Time: Monday, August 8, 2022 : 8:30 AM to 10:20 AM
Sponsor: ENAR
Abstract #319188
Title: Tensor Reduced-Rank Regression with Incomplete Observations, with Application to Longitudinal Microbiome Analysis
Author(s): Gen Li*
Companies: University of Michigan
Keywords: tensor array; reduced-rank regression; microbiome; log-contrast model; ADMM
Abstract:

Multivariate longitudinal data measured on a regular grid can be concisely represented as a three-way tensor array (i.e., sample-by-feature-by-time). Missing observations are commonly encountered in such data since not all samples are measured at every time point. The missing data impose significant challenges for statistical analysis. This work develops a novel scalar-on-tensor regression framework, called TRIO, which effectively leverages all available observations in a design tensor for accurate parameter estimation and prediction. We propose a parsimonious model for the design tensor and regression coefficient matrix and devise a computationally efficient algorithm to estimate model parameters with flexible regularization. Numerical studies demonstrate the superior performance of the proposed method over competitors. The method is further applied to a preterm infant study to predict neurodevelopment response from longitudinal microbiome data.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2022 program