Incomplete generalized U-statistics for food risk assessment.

Area of Science:

• Food safety and toxicology
• Statistical modeling
• Risk assessment

Background:

• Assessing health risks from food contaminants requires robust statistical tools.
• Estimating exposure probability against established intake limits (e.g., PTWI) is crucial for public health.
• Independent consumption and contamination data present unique analytical challenges.

Purpose of the Study:

• To develop and evaluate statistical methods for quantifying food contaminant risks.
• To estimate the probability of exceeding provisional tolerable weekly intake (PTWI) levels.
• To propose and validate methods for constructing confidence intervals for risk estimation.

Main Methods:

• Utilized a Monte Carlo approximation of a plug-in estimator, related to generalized U-statistics.
• Investigated asymptotic properties of the proposed estimator.
• Developed and compared two asymptotic variance estimators: bootstrap and approximate jackknife, leveraging Hoeffding decomposition.

Main Results:

• The study provides asymptotic properties for the proposed statistical estimator.
• Several confidence intervals for risk estimation were proposed and evaluated.
• The methods were illustrated with a case study on Ochratoxin A exposure in France.

Conclusions:

• The proposed statistical tools offer a quantitative framework for evaluating food contaminant risks.
• The methods are applicable when independent consumption and contamination data are available.
• The study contributes to improved risk assessment strategies for food safety.