How should non-detects be used in risk assessment?
DEQ recommends the use of the ProUCL software available through EPA’s website at http://www.epa.gov/esd/tsc/software.htm for statistical analysis of data used for risk assessment purposes. The most recent version of the ProUCL software is version 4.00.04, an upgrade of Pro UCL version 4.0, released by EPA in 2007. The ProUCL version 4.00.04 software provides statistical analysis of non-detects or censored data; therefore, there is no need to use half the detection limit for undetected compounds (with the exception of calculations of toxicity equivalence quotients for dioxins/furans or other dioxin-like compounds – please refer to the FAQ regarding sampling for dioxins/furans for more information). The ProUCL software also provides information regarding whether there is enough data available for statistical analysis, whether the data set falls into a particular distribution, whether an upper confidence limit (UCL) other than 95% (i.e., 97%, 99%) is appropriate given the data set; and al
DEQ recommends the use of the ProUCL software available through EPA’s website at http://www.epa.gov/esd/tsc/software.htm for statistical analysis of data used for risk assessment purposes. The ProUCL software provides statistical analysis of non-detects or censored data; therefore, there is no need to use half the detection limit for undetected compounds (with the exception of calculations of toxicity equivalence quotients for dioxins/furans or other dioxin-like compounds – please refer to the FAQ regarding sampling for dioxins/furans for more information). The ProUCL software also provides information regarding whether there is enough data available for statistical analysis, whether the data set falls into a particular distribution, whether an upper confidence limit (UCL) other than 95% (i.e., 97%, 99%) is appropriate given the data set; and also provides a limited outlier analysis. It is generally appropriate to follow the ProUCL recommendations. Thorough justification and DEQ approv
