MP8 Current Trends in Toxicity Evaluation
Room 19A/B, Level 4
2:10 PM - 5:30 PM, Monday, 10 November 2003
Chair: Ownby, David ,
Co-chair: Newman, Mike ,
(147) Predicting Sediment Toxicity Using Logistic Regression.
smith, e1, 1 Department of statistics Virginia Tech, Blacksburg, VA, 24061
ABSTRACT- Logistic regression was used to evaluate the ability of information on the chemical concentrations in sediments to correctly predict the results of a toxicity test for the sediment. Using a data set of matched sediment toxicity and chemical data for a number of studies, several approaches for predicting toxicity based on multiple logistic regression with concentration addition models were investigated. Two models were selected as meeting criteria for acceptability: The first uses individual chemicals selected using stepwise selection. The second uses derived variables to reflect combined metal contamination, polycyclic aromatic hydrocarbon (PAH) contamination and the interaction between metals and PAHs. Overall, the models suggest that toxicity may be correctly predicted around 77% of the time although prediction is better for not toxic samples than for toxic samples.
Key words: Statistical Methods, Sediment Toxicity, Concentration addition, Bioassay