A random subset implementation of weighted quantile sum (WQSRS) regression for analysis of high-dimensional mixtures

Paul Curtin, Joshua Kellogg, Nadja Cech, Chris Gennings

Research output: Contribution to journalArticlepeer-review

32 Scopus citations

Abstract

Here we introduce a novel implementation of weighted quantile sum (WQS) regression, a modeling strategy for mixtures analyses, which integrates a random subset algorithm in the estimation of mixture effects. We demonstrate the application of this method (WQSRS) in three case examples, with mixtures varying in size from 34 to 472 variables. In evaluating each case, we provide detailed simulation studies to characterize the sensitivity and specificity of WQSRS in varying contexts. Our results emphasize that WQSRS is robustly effective in evaluating mixture effects in diverse high-dimensional contexts, yielding sensitivity and specificity in empirical contexts of approximately 73–75% and 73–89%, respectively.

Original languageEnglish
Pages (from-to)1099-1114
Number of pages16
JournalCommunications in Statistics Part B: Simulation and Computation
Volume50
Issue number4
DOIs
StatePublished - 2021

Keywords

  • Mixtures models
  • Weighted quantile sum regression
  • mixtures
  • supervised learning

Fingerprint

Dive into the research topics of 'A random subset implementation of weighted quantile sum (WQSRS) regression for analysis of high-dimensional mixtures'. Together they form a unique fingerprint.

Cite this