linear, square-root and log transformations. The data and fitted curves are plotted on the log scale, so that the fitted log model appears linear and the linear model shows the highest degree of curvature. What becomes clear from Figure 7-5 is that variations in estimated BMDs are not explained by differences in how well the models fit the bulk of the data but rather what the models predict for the mean response for unexposed individuals.

Figure 7-4 The estimated expected excess response due to Hg exposure as a function of the Hg concentration calculated using the linear, square root, and logarithmic (log) model. Source: E. Budtz-Jørgensen, Copenhagen University, N. Keiding, Copenhagen University, and P. Grandjean, University of Southern Denmark, unpublished material, May 5, 2000.