I completed a doctorate in Epidemiology at Emory University in 2004. My dissertation, titled "Hierarchically well formulated models in epidemiology," examined hierarchically well formulated models (HWF) in the context of epidemiology. Notably, HWF models are invariant to linear transformations of the independent variables. Peixoto proved this for the special case of multiple regression (linear) models in 1990. I used a more general proof that extends this property to generalized linear models. I also developed a uniform method for computing epidemiologic effects from a model regardless of its type (linear, logistic, poisson, etc.). This method introduces an intermediate result that I call "generalized effects." I believe these are useful as a way of approaching classes of models and for didactic purposes.
Alternatively, you can download the PDF file I created to print my dissertation. The print quality of this file is much better, it is a smaller file, and the page numbers of the PDF match the page numbers of the dissertation.
Unfortunately, neither version is searchable. My dissertation was prepared in TeX and converted to a PDF file for printing. I do not know if there is a way to produce a searchable PDF version.
In the second graph in example 38 (P. 64), the label for the vertical axis appears as "f x", but should be "f(x)".