Abstract:

The mercury exposure in milled rice has become a hotspot in risk assessment. Considering the positively skewed contamination data, an economic and effective assessment must be based on an appropriate sample size. With the benefit of computer simulation, this paper set out to study the relationship between sample size and percentile estimate with Weibull distribution, which was used to fit the mercury contamination data. The simulation results showed that accurately estimating a higher percentile of positively skewed distribution, specially the one on the right side of peak, would require a larger sample size. And a larger sample size always resulted in a more accurate and more stable estimated percentile. As sample size increased to exceed 400, 700 and 1500, estimates of P95, P97.5 and P99 could obtain guaranteed accuracy and precision respectively, while a sample size of 1500 could only satisfy the accuracy but not precision of P99.9 estimates. Furthermore, insufficient samples would always lead to an underestimate of high percentile, and then lead to an underestimate of risk. When it comes to risk assessment of mercury exposure in milled rice, the in-depth study of sample size promises not only a possibility of an accurate result, but also an economical resource allocation.

Key words: milled rice, mercury, exposure assessment, Weibull distribution, percentile, sample size