A latent Gaussian model for multivariate consumption data

D.J. Allcroft, C.A. Glasbey, M.J. Paulo

Research output: Contribution to journalArticleAcademicpeer-review

9 Citations (Scopus)

Abstract

We propose a multivariate statistical model for individual consumption of multiple food types, to provide a more objective basis for exposure assessment from chronic consumption. Intake of each type of food is modelled by a latent Gaussian variable, where intake is zero if the latent variable is below a threshold, and otherwise is a monotonically increasing function of the latent variable. Further, we use a Factor Analysis model to describe the association in intakes between different foods. This reduces the number of parameters to be estimated and aids interpretation. The method is illustrated using data from the Dietary and Nutritional Survey of British Adults, 1986¿1987.
Original languageEnglish
Pages (from-to)508-516
JournalFood Quality and Preference
Volume18
Issue number3
DOIs
Publication statusPublished - 2007

Keywords

  • exposure assessment
  • pesticides
  • chemicals
  • food

Fingerprint Dive into the research topics of 'A latent Gaussian model for multivariate consumption data'. Together they form a unique fingerprint.

Cite this