Putting the G in GAM

[dschoenig]

Currently the use of other distributions is treated rather late in the workshop. I think it would be beneficial move this section forward (i.e. before talking about different basis functions), and show how the same concepts as for GLMs apply. By referring to the GLM course, this section could also be much shorter (one example plus challenge).

Now that we are using the ISIT data set, this can be integrated readily, as the Gaussian fit actually violates model assumptions (variance homogeneity). Given that the response can be modeled with a distribution from the Tweedie family , it would be straightforward to introduce the Generalized.

The model checking part mentioned #13 would be a natural intro to the generalized section.

[dschoenig]

3142f2e replaces the previous section on "other distributions" with a section "Generalizing the additive model", now coming after the interaction part.