MA3 WET Testing
Ballroom G, Level 4
8:00 AM - 12:00 PM, Monday, 10 November 2003
Chair: Hall, Scott , Gully, Joe ,
(023) Whole Effluent Toxicity Tests: Using Bayesian Methods To Calculate Model-Based Endpoint Variability.
Warren-Hicks, W1, Parkhurst, B1, 1 The Cadmus Group, Chapel Hill, NC, USA
ABSTRACT- As part of this recently completed study for the Water Environment Research Foundation (WERF), innovative statistical procedures are employed to calculate the precision of model-based endpoints derived from a national data base of whole effluent toxicity (WET) tests. Variance components for both model-based endpoints like the IC25 or LC50 and biological endpoints were calculated. The mathematical framework used to calculate the variance components is called a hierarchical Bayesian model. This framework is useful because information about toxicity can be partitioned into several levels (similar to a decision tree). At each level, estimates of uncertainty can be calculated. The levels are not independent, the information at lower levels has a direct influence on the calculations at the next level. Thus, the framework establishes a mathematical dependency (or in statistical terms, a conditionality) among the various hierarchies. The findings from this study have wide implications on the use of WET tests in regulatory decisions, like the finding of a reasonable potential (RP) for toxicity, and compliance determinations. This paper presents the statistical methods and associated variance calculations. The companion paper in this session presents a case study using the calculated variance components.
Key words: uncertainty, WET, effluent, variability