Analysis of dichotomous response data from toxicological experiments involving stable laboratory mouse populations

The comparability between an analysis of variance that employs untransformed proportions and one that employs two common arcsine transforms is assessed by using preweaning mouse litter survival data from a toxicological experiment that involves stable laboratory populations. The importance of controlling for litter order, litter size and birth cohort is demonstrated. An adaptation of a nonparametric method suggested by Mantel (1963, Journal of the American Statistical Association 58, 690-700) is proposed as an alternative procedure. Although the results indicate that the nonparametric method is modestly less powerful than the parametric approach, its use is recommended when the assumptions underlying the analysis of variance are questionable. A method for plotting the temporal pattern of the effect of continuous exposure to the toxicant (DDT) on preweaning survival is also presented.