Homework

Listed below are the assigned homework problems. Although these homework problems will NOT be turned in for a grade, you should expect to see similar problems or these exact problems on tests and projects. You will be at a great disadvantage when taking tests and completing projects without doing these homework problems. Partial answers with R code and output are available for some homework problems.

Chapter 1

Run the R program HS_college_GPA.R. The data set needed for it is located on the data sets webpage.

1.20, 1.24, 1.28, 1.45

	Partial answers
	Note that all data sets in KNN are on the CD accompanying the textbook. If your book did not come with it, go to the data sets web page to download them.

Page 23-24 proofs

Chapter 2

	2.2, 2.5, 2.11, 2.12, 2.14 (replace d. with: use R to create a scatter plot with the estimated regression line, 90% C.I., bands, and 90% P.I. bands), 2.24 (only a.-c., do not do the Table 2.3 part), 2.30, 2.31 (a.-c.), 2.64.
	Partial answers

Chapter 3

	3.4 - Skip b. For a, use box and dot plots. For c, use a histogram. For e, use a QQ-plot with a normal distribution and skip the correlation part. For g, also use Levene's test.
	3.8 - For a, use box and dot plots. For b, use a histogram of the residuals or semistudentized residuals. For d, use a QQ-plot with a normal distribution and skip the correlation part. For e, use the median of X instead of 69.
	3.27 - use examine.mod.simple() to examine the models in part a.
	Reproduce the Box-Cox transformation results for the example on p. 136. See the partial answers below for more on this problem.
	Partial answers
	Old STAT 870 project: Diamond_data_set.doc, diamond_prices.xls

Chapter 4

	4.3 (only b.-c.), 4.7 (only a. and c. using the Bonferroni procedure), 4.27 (only a.-b. and d. using the Bonferroni procedure), and the extra problem in the partial answers document (data set: SO2.DAT)
	Partial answers

Chapter 5

	5.4, 5.12, 5.23
	Find the following for the copier maintenance data set (see #1.20) using matrix methods (do not use lm()): b, Y^, e, SSTO, SSE, Cov^(b), 95% C.I. for E(Y_h) at X_h=6, and 95% P.I. for Y_h(new) at X_h=6.
	Partial answers

Chapter 6

6.15(a.-e., g.)

	For a: Do box and dot plots instead and include Y
	For d: Skip
	For e: Use the semi-studentized residuals; do not examine the two-factor interaction; include a histogram with normal distribution overlay
	For f: Skip
	For g: Do Levene's test as well (once for each predictor variable - there is not a good way to divide the data)
	Also, perform the Box-Cox procedure and comment on its results
	You may use examine.mod.multiple.R to help answer this problem

6.16, 6.17

6.30

	For a: Do box and dot plots instead
	For d.: Also calculate and interpret the adjusted R²
	For e.: Don't do the two factor interaction plots
	You may use examine.mod.multiple.R to help answer this problem.
	Supplementary problems (download Word file)

Partial answers

Chapter 7

	7.5, 7.6, 7.14, 7.26, 7.38
	Partial answers

Chapter 8

8.38

	For part a., use examine.mod.multiple.R to examine the model
	Also, construct a scatter plot of the data with the second-order sample model plotted upon it. Use this to help justify your results in c.

8.40

8.41

Partial answers

Chapter 9

9.10

For part a., use examine.mod.multiple.R to examine the model

9.11

9.18

a. Use AIC and hypothesis test methods

9.21

The last sentence refers to material in Section 9.5 (wait until this section to answer the question)

9.25

	For part a., use examine.mod.multiple.R to examine the model
	For part c., see p. 358 for what KNN refers to as "bias" (paragraph below equation 9.10)

Partial answers

Chapter 10

10.7, 10.11, 10.17

10.27

	Use examine.mod.multiple.final.R to help answer the questions
	b: You do not need to do the formal normality test
	d: Replace the dot plot with the plot created by examine.mod.multiple.final()

9.27 (model validation problem)

Partial answers

Chapter 11

Reproduce the example on p. 427-9 of KNN

	KNN uses my "method #4" for obtaining the weights
	Try putting the data into 5 groups to obtain weights also; compare these results to those obtained by KNN

11.11

Instead of b.-e., use LMS and LTS to fit the model and plot it; compare the estimates to least squares

11.25

	b: Use a 3D plot instead of a contour plot; examine a full second-order model as well
	c: Use loess() with q = 9/25
	d: Use a 3D plot instead and make sure to put it on the same scale as in b for comparison purposes

Partial answers