Abstract
ABSTRACT
Risk assessors typically estimate long‐term exposure by calculating a 95th percentile upper confidence level (95UCL) on the mean concentration across an exposure area. There are several methods for calculating a 95UCL. The United States Environmental Protection Agency has developed software (ProUCL) that calculates and recommends a 95UCL methodology based on the properties of the dataset. Release of ProUCL Version 5.2 in 2022 demonstrated a significant change in the decision tree used to recommend a 95UCL method. To evaluate the differences between ProUCL 5.1 and 5.2, a simulation study was performed that includes a range of samples sizes, coefficients of variation, and frequency of non‐detects often seen in environmental datasets. Over 160,000 datasets were evaluated, and 95UCLs were calculated using both versions of the software. The results demonstrate a trade‐off in coverage and accuracy between the two versions. ProUCL 5.2 recommends 95UCLs that exhibit greater accuracy but have lower coverage. ProUCL 5.1 recommends 95UCLs that achieve higher coverage, but accuracy is typically reduced by a factor of two. Future versions of ProUCL software could be improved through revisions to the decision tree that provide an appropriate balance between coverage and accuracy while also allowing advanced users to evaluate alternative performance criteria within applicable regulatory requirements.