
This assignment is due in class on 
Wednesday, March 8. It requires
substantial computation; start early.

I talked about "gamma=(C-D)/(C+D)" last week;
Agresti introduces it in subsection 2.4.4
and describes its advantages for ordinal variables
in subsections 3.4.3 and 3.4.4. 
(If you don't know what Agresti means by "power," 
look up the definition of "statistical power." 
It's even on Wikipedia.)

The attached Goodman-Kruskal paper,
cited by Agresti, goes into
greater depth about gamma.
Its subsection 3.5 explains how to estimate 
a 95% confidence interval for gamma.

Group 1: Edward, Raymond, Noemi, and Analia,
for Agresti's Table 10.5, estimate
a 95% confidence interval for gamma.

Group 2: Aracely, Gamaliel, and Roberto,
for Agresti's Table 3.2, estimate
a 95% confidence interval for gamma.

Don't try to read the whole G-K paper; 
focus on what you need to know 
for computing the confidence interval.
Their example computation of a
gamma confidence interval is extremely
helpful.

Back in 1963, Goodman and Kruskal computed
their example gamma confidence interval
"rapidly with the aid of a table of squares and
a desk computer," but still found it "rather tedious."
Depending on the skills of your group,
you may or may not find it helpful
to automate some parts of the computation 
using Excel or even a programming language.
