Estimating the logistic curve from grouped data

Abstract

The problem of estimating the parameters of the logistic dose-response curve from data consisting of single binary responses at each of several distinct doses isdiscussed. Four estimation methods are considered; three methods require grouped data and thefourth is the m.1. estimator computed from the original, ungrouped data. Two of the methodsrequiring grouped data are based on the minimum logit chi-square technique. The results, based on monte carlo simulation , are: (I) the grouped data methods often yield smaller mean squared errors when estimating α and β (2) the m.1. estimators yield smailcr mean squared errors than the minimum logit chi-square estimators when estimating to 50% or 90% effective dose.