Stat 665 (Spring 2008)
Kaizar
Homework 4
Due at the beginning of class, Monday April 28.
Aim of Homework: To select good models.
Reading: An Introduction to Categorical Data Analysis, 2nd edition: Chapter 5
Exercises:
1. Interactions without Main Effects (3 points) The following is a prediciton equation for
predicting the effects of temperature (in degrees Celsius) and pressure (in mm Hg) on the log
odds of rain:
logit(πrain) = α+βX1X2
(a) Write down the prediction equation when you change temperature from degrees Celsius
to degrees Kelvin. (Degrees Kelvin = Degrees Celsius + 273.15)
(b) Based on the previous part, explain one reason why it is a good idea to include main
effects in any model that includes those effects in interactions.
2. Prediction as Fit Criterion (5 points)
Agresti problem 5.10, parts a, b, d, and e
3. Model Selection (5 points)
Agresti problem 5.13. Note that the data is in a file that is separated by spaces, so if you are
using R, don’t include an option like sep=”,”.
4. Evaluating Model Fit (5 points)
Appropriately evaluate the fit of the model you chose in the last problem. Be sure you comment
on deviance, residuals and influence.
5. Summarizing Your Results (4 points)
Write three paragraphs regarding your credit scoring model. One of these should explain how
you searched for your model and why you chose that methodology. One should explain the
results of your evaluation of the model fit. One should interpret your model in the context of
the problem.
6. Bits and Pieces (3 points)
Agresti problem 5.30. Be sure to explain your answers.
7. Group Direction (0 points) On a separate sheet of paper not attached to the rest of your
homework, write 2-3 sentences describing what you think the topic of your project will be. Be
sure all group members’ names
1
