Laserfiche WebLink
dilution-water controls would show normal receptors. To reduce <br />potential for investigator bias, SEM surveys of olfactory <br />epithelium of control and toxicant-exposed fish were limited to <br />the same general regions of the olfactory surface. The target <br />region was systematically surveyed at a magnification of 10,000X <br />for a period of 15 min. The number of ciliated and microvillous <br />receptor cells that were encountered during the survey period was <br />recorded. At least three fish from inhibitory toxicant <br />concentrations and controls were examined. <br />Statistical Analysis <br />Linear logistic regression by the method of maximum <br />likelihood (SAS Institute 1990), was used to analyze the binomial <br />response variable for olfactory ability (inhibited or not <br />inhibited) as a function of toxicant concentration. <br />Intercept-only, and intercept-and-covariate models were evaluated <br />for each analysis and the amount each parameter improved a model <br />was assessed with AIC. Goodness of fit of selected models was <br />assessed by calculating a chi-square statistic for model <br />deviance (SAS Institute 1993), and examining data and residual <br />plots. Prior to regression analysis, toxicant concentrations <br />were loglo transformed. Regression equations were used to <br />estimate median effective concentrations (EC50) and their 95% <br />confidence intervals for inhibition of olfaction. Concentrations <br />of copper and zinc estimated to inhibit olfaction in 1% of the <br />test organisms (EC1) were also calculated. The 1% level was <br />9 <br />