Interval estimation of the ED50 when a logistic dose¯response curve is incorrectly assumed

For a correctly assumed logistic dose¯response curve, a number of authors have considered the various interval estimation methods for the median effective dose (ED50). In this paper, we look at the performance of four large sample parametric methods and four bootstrap resampling approaches for the problem of constructing a confidence interval of the ED50 and investigate the robustness of these methods when a logistic dose¯response curve is incorrectly assumed. The underlying true dose¯response curve studied is, respectively, either the probit model, the cubic logistic model or the asymmetric Aranda-Ordaz model.