Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 485 - A Unified View on Model Selection
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: Biometrics Section
Abstract #313649
Title: Estimating Overall Contribution of High-Dimensional Predictors with Measurement Error
Author(s): Jianxin Shi* and Soutrik Mandal and Do Hyun Kim
Companies: National Cancer Institute and National Cancer Institute and Biostatistics Branch, DCEG, NCI
Keywords: overall contribution; heritability; microbiome; measurement error

High dimensional genomic profiling studies, including DNA methylation, transcriptome sequencing and microbiome, are frequently performed in prospective studies to identify risk factors for disease risk or clinical outcomes. For a complex phenotype, there might be many independent factors, each of which has only modest effect size. To successfully design a study to identify risk factors and to build predictive models, one has to estimate the overall contribution from all predictors. The linear mixed model developed for estimating the heritability in genetic studies can be used for this purpose. However, the measurement error in the genomic predictors may complicate the estimate of the overall. We here develop a statistical method for unbiasedly estimating overall contrition with accuracy verified by extensive simulations. The algorithm will be exemplified in a prospective study investigating the relationship between human oral microbiome and cancer risk.

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

Back to the full JSM 2020 program