Required Homework #2:

1. Modify the in-class R to fit a logistic function (in place of the normal
ogive) to the Bock & Jones Constant Method Saltiness data. Do this for both the 
derivative-free version and the version with derivatives. [Hints: If the latter 
seems difficult, reconsider some of the things we said about logistics in class 
last week. It should not be difficult.]

2. Modify the in-class C++ to fit a logistic function (in place of the normal 
ogive) to the Bock & Jones Constant Method Saltiness data.

In both of the above, hand in your program code, the results you obtain, and
a short report on the process. Which model fits better, the normal or the
logistic?

This homework is due in class on Monday, November 3.





Optional Homework Exercises, Set #6, for Bock & Jones class #1:

A. Augment the report, above, with a compare-and-contrast exercise between 
Thurstone's Model and the model that yields  the logistic as the function being 
fitted (hint: That's usually called "Luce's Choice Model").

B. Figure out (from the references in the R Help for glm, or from other
sources) just exactly what R's glm function is doing when it says it's 
doing "iteratively reweighted least squares," and provide a report on
that to the rest of us. Is it, indeed, exactly the same as ML?