5.6. Refer to Table 2.5. To investigate the association, one can use a logistic regres- sion model, but interchange roles of response and explanatory variables.

a. Fit the model, treating party identification as nominal. Note that it is satu- rated. (The likelihood-ratio test for an absence of a party effect in this model is identical to the G test of independence in Section 2.4.)

b. Fit a model that treats party identification as ordinal with equally spaced scores, and use the model to test independence. Interpret. Show that the model implies that the odds ratio for the first two columns is identical to that for the last two columns. Report this estimate, and interpret.