Comparison of probit analysis versus arcsine square root transformation on lc50 estimation

Abstract

Listen

Translate article

The probit analysis (PB) method, the linear regression of probit mortality on log dosage, has long been employed to estimate lc 50. However, when using the PB method, cases of concave distribution of residuals with respect to the predicted line after fitting, a violation of linear regression principle, have been found to often occur. Under these circumstances, an alternative method of regressing the arcsine square root of mortality against dosage (ARS) has been found not only to be more effective in transformation for linearity but also to have a higher insignificance level in lack-of-fit test than the PB method. lc 50 obtained by the PB method is smaller than that by the ARS method. The advantages of ARS over the PB method are: a statistically sound estimate of lc 50; the possibility to adjust lc 50 by natural mortality; and the ability to incorporate dose-response and environmental conditions into a model. The lc 50 discrepancies between these two methods are quantified and remedies for the existing overestimate of lc 50 by the PB method are provided.