Probit Analysis of Correlated Data: Multiple Observations Over Time at One Concentration

Abstract

We present a method to analyze serial time–mortality data from bioassay experiments where successive observations are made on the same group of organisms exposed to 1 concentration of a stimulus (for example, a pesticide). Standard probit techniques are not applicable because such data are correlated. Methods are presented for calculating statistics for the time–mortality line by regressing complementary log-log, logit, or probit transformations of proportion of responders on untransformed time or logarithmic transformations of time; and for calculating confidence limits on lethal time values corresponding to given mortality levels. As a result we have developed a computer program to facilitate use of the method.