Sensitivity analysis for high percentiles of Ochratoxin-A exposure distribution - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2001

Sensitivity analysis for high percentiles of Ochratoxin-A exposure distribution

Résumé

High percentile estimations of the exposure to the mycotoxin Ochratoxin-A (OTA) in food, for the French population, were calculated in a previous study by a Monte Carlo type simulation method from real consumption and contamination data [1]. In this paper, from the same data (but only for the children class), we focus on sensitivity analysis (SA) of the high 95th and 99th percentiles (the simulation outputs), relatively to the variation of the parameters of the fitted probability density functions (the simulation inputs), necessary for having a relevant and stable estimation of these percentiles. After some preliminary trials, we postulated a quadratic polynomial model and we used an experimental design approach depending on a resolution-V fractional factorial design of 6561 experiments to lead to an optimal estimation of the polynomial model parameters. The factors ranges were established by bootstrap sampling taking into account the consumption dependencies by the Iman & Conover method [2] and, eventually, taking into account the parameter correlation of the fitted probability densities. Finally, we have validated and useful parsimonious polynomial models for each desired percentile showing a major influence of the distribution parameters of the two foods « Cereals » and « Pork », and eventually three with « Fruit Juices », in the sensitivity of the percentiles
Fichier non déposé

Dates et versions

hal-01263594 , version 1 (27-01-2016)

Identifiants

  • HAL Id : hal-01263594 , version 1
  • PRODINRA : 36052

Citer

Isabelle Albert, Jean-Pierre Gauchi. Sensitivity analysis for high percentiles of Ochratoxin-A exposure distribution. International symposium on sensitivity analysis of model output, Jun 2001, Madrid, Spain. ⟨hal-01263594⟩

Collections

INRA INRAE MATHNUM
30 Consultations
0 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More