Stat 5302 Spring 2011

R. D. Cook

Assignment 10

Due in Lab Thursday, April 28

Reading: Chapters 13, 14 and 15.

Problems to be handed in: 13.3, 14.3.3, 14.6, and problems 1, 2, 4, 5, and 6 below.

1. Assuming that the log variance is a linear function of the mean, calculate the iteratively re-weighted least squares estimator of the coefficients in the transaction data, using three sets of weights (iterations) constructed as discussed in class.

2. Why is transforming the response and predictors simultaneously to multivariate normality desirable?

The following problems refer to the regression of log(BigMac) on the 4 terms log(Bread), log(BusFare), log(TeachSal) and log(TeachTax) in the Big Mac data (big-mac.lsp).

3. Assuming a constant variance function, fit the mean function that is linear in the 4 terms and find the residual and the leverage for Lagos. Use these values to compute by hand the Studentized residual, the outlier t value and Cook's distance for Lagos. Compare your results with those given by Arc.

4. Suppose you are about to vacation in Mexico City and would like to know if the cost of BigMacs will be unusually large. Test the hypothesis that Mexico City is an outlier. Explain what the test means (eg. what is an outlier?).

5. Find the case with the largest absolute outlier t value and test that it is an outlier.

6. Find the case with the largest Cook's Distance. What test do you use to assess the significance of Cook's distance? Is the case influential because it is a high leverage case, an outlier or a combination of these possibilities? Delete the case and briefly summarize its impact on the analysis.