Resolution of additive mixtures into source components and contributions: A compositional approach

Research output: Contribution to journalArticlepeer-review

Abstract

Methodology is developed for analysis of observations that are random linear combinations of point “source components.” Dual goals are to estimate unknown source identities and to characterize the mixing process by which sources contribute to the observations. Observations are modeled as arising from a mixture distribution, whereby the mixing component characterizes the process of interest and the kernel component captures measurement error. A parametric model is proposed, and maximum likelihood estimates of source and mixing parameters are obtained. Estimate performance is investigated by Monte Carlo simulation. Major results are devoted to studying a constraint framework within which model identifiability is guaranteed. For maximal generality, a compositional framework is applied throughout. The resolution problem discussed in this article is common in the physical sciences. For illustration, an application to air pollution data is presented.

Original languageEnglish (US)
Pages (from-to)1450-1458
Number of pages9
JournalJournal of the American Statistical Association
Volume89
Issue number428
DOIs
StatePublished - Dec 1994

Keywords

  • Air pollution
  • Identifiability
  • Latent variables
  • Measurement error
  • Mixture model
  • Source apportionment

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Fingerprint Dive into the research topics of 'Resolution of additive mixtures into source components and contributions: A compositional approach'. Together they form a unique fingerprint.

Cite this