ABSTRACT
Thus far we have illustrated the use of linear regression as a way of fitting dose-effect data. When the effect is plotted against the logarithm of the dose over some mid-range of doses, a somewhat linear trend often results. In this limited dose range, the points are often well approximated by a straight line fit; thus, if there are sufficient points in this linear region, the use of linear regression is often adequate to get
D
values and their standard errors. However, the complete log dose-effect curve, that is, the curve over an extensive range of log dose values, is frequently sigmoidal (S-shaped) and, thus, nonlinear.
