Stat 5302 Spring 2011

R. D. Cook

Assignment 11

Due in Lab Thursday, May 5

Reading: Chapter 21, 22

Problems to be handed in:

The following problems all refer to the data on recumbent cows in file downer.lsp

21.11 Consider the regression of Outcome on CK alone. Construct histograms of CK and log(CK), each marked by Outcome, and smooth each by Outcome with a Gaussian Kernel Density estimate with tuning parameter 1. Based on these plots alone, would it be sound to use log(CK) rather than CK to form U-terms in a logistic regression model? Does the logisitc model h₀ + h₁ log(CK) seem reasonable for this regression? Provide brief justifications for your responses. (See Section 22.1)

21.12 Provide a careful interpretation of the estimated coefficient of log(CK) in the regression of Outcome on log(CK). That is, what does the estimated coefficient tell us about the relationship between Outcome and log(CK)? You will likely want to use conditional distributions as part of your answer.

21.13 Construct a scatter plot of log(AST) versus log(CK) marked by Outcome. Imagine fitting the logistic model h₀ + h₁ log(CK) + h₂ log(AST). Draw an estimate of

h₀ + h₁ log(CK) + h₂ log(AST) on the plot (See Section 22.2).

21.14. Study the regression of Outcome on the four predictors AST, CK, PCV, and Inflamat. Prepare your solution in two parts. Part 1 is to be an at most 400 word summary of your findings. Part II should consist of at most two pages of supporting material ( computer output, graphs, ...) to support your findings in Part I. You may refer to Part II when reporting your findings in Part I.

The exam data for the final will be given on May 4. There will be no take-home final exam.