Nonparametric bayesian model to cluster individual co-exposure to pesticides found in the French diet.

Amelie Crepet (1) , Jessica Tressou (2)

1 AFSSA - Agence Française de Sécurité Sanitaire des Aliments
2 MET@RISK - Méthodologies d'Analyse de Risque Alimentaire

This work introduces a specic application of bayesian nonparametric statistics in the food risk analysis framework. The goal is to determine cocktails of pesticide residues to which the French population is simultaneously exposed, so as to give directions for future toxicological experiments for studying possible combined effects of those cocktails. For that, the joint distribution of exposures to a large number of pesticides, called the co-exposure distribution, is assessed from the available consumption data and food contamination analyses. We propose to model the co-exposure by a Dirichlet process mixture based on a multivariate Gaussian kernel so as to determine groups of individuals with similar co-exposure patterns. The study of the correlation matrix of these sub-populations will permit to dene the cocktails of pesticides to which they are jointly exposed at high doses. The posterior distributions and the optimal partition are computed through a Gibbs sampler based on stick-breaking priors. To reduce computational time due to the high dimension of the data, a random block sampling is used. As an extension, we propose to account for the uncertainty of food contamination through the introduction of an additional level of hierarchy in the model. The results of both specifications are exposed and compared.

Mots clés: food risk analysis; Dirichlet process; nonparametric bayesian modeling; multivariate normal mixtures; clustering; multivariate exposure; food risk analysis.

Domaines: Statistiques [math.ST]; Théorie [stat.TH]

Amelie Crepet, Jessica Tressou. Nonparametric bayesian model to cluster individual co-exposure to pesticides found in the French diet.. 2009. ⟨hal-00438796v1⟩