Selected Errata and Notes from Hair et al's 5th Edition

Errata and Notes on Multivariate Data Analysis, 5th edition

Errata:

Page Correction

20 Figure 1.2: in linear probability models chapter 4 should be chapter 5

28-29 Table 1.3: reference to indep/dep is missing from Table

42 the 5th line from the bottom, should say "skewness in the opposite direction is indicated."

72 Figure 2.7: graph axes are reversed relative to SAS and to p81

74 Figure 2.8(b): V1 distribution should be positively skewed

81 Figure 2.11: z should be |z| (twice)

90 Glossary entry for Reliability: remove "is intended to" and make "measure" plural

171 b_total should be b1_total

189 3rd paragraph: Table 4.9 is on page 192, not 193.

191 Exhibit1: r_y,x1(x2) eqn should have 60- in the numerator

191 Exhibit1: second r_y,x1(x2) should be r_y,x2(x1)

191 Exhibit1: r_y,x2(x1) should refer to x_2 = .11^2, not x_1

225 Line 13: +-2sqrt(n), should be +-2/sqrt(n)

229 3rd line from bottom: value should be 0.08, not 0.2

244 Expressions regarding Zjk should refer to Wij, not Wi

193 8th line: VIF value should be 10, not .10

195 Line 4: higher should be lower

199 First line in R Square section: expression should read (.7012 = .491)

201 4th line from bottom: value should be 0.014, not 0.14

241 Under "centroid" other uses of centroid (other than for Z scores) are also common. For example, the vectors of variable means for the original variables - when segregated into groups - are referred to as group centroids. The textbook prefers the name group mean as used in Table 5.1 on page 248.

241 Under "discriminant function," all W symbols should have a second subscript j for the jth discriminant function. For example, W_ij or W_ji. Also the intercepts should also be indexed with j.

245 in Figure 5.1 each figure should have the left-hand symmetrical area shaded also

290 Two corrections:

6th line: correctly classified is 85.0 percent, not 86.4%

Press's Q calculations: should use 55 and 34 rather than 56 and 37. This results in new Q values of 41.7 and 19.6

291 Make the following changes in Table 5.10

The second (extra) line containing "Total Value Analysis" with values of 38, 4 and 34 should be deleted.

The percent of grouped cases correctly classified for the holdout sample is 85%.

335 4th line above T^2 equation: "N1+N2-2-1" should be "N1+N2-p-1"

Notes:

29 dummy coding for "specification buying" is unintuitively assigned

99 SAS does in fact supply partial correlation matrix

292 Add following footnote at bottom of Table 5.11 Note: Cases 82 and 64 are predicted to be in Group 0 rather than Group 1 since SPSS utilizes standardized discriminant scores when separate variance estimates are used

Chapter 5 Note on classification process: Because separate variance estimates were empolyed in the two group example, SPSS utilizes standardized Z scores in classification. This results in two cases (#82 in the analysis sample and #64 in the holdout sample) to be classified as Group 0 (Specification Buying) whereas the simple cutting score approach would classify them as Group 1 (Total Value Analysis)

The cutting score approach described in the text is a simplified description of the classification process and may not exactly replicate actual results in the instances of separate variance estimates, use of Fishers's linear discriminant functions for classification or other options. It does, however, accurately describe the basic process and will result in identical results in most situations.

Go back to DSc8450 home

Page	Correction
20	Figure 1.2: in linear probability models chapter 4 should be chapter 5
28-29	Table 1.3: reference to indep/dep is missing from Table
42	the 5th line from the bottom, should say "skewness in the opposite direction is indicated."
72	Figure 2.7: graph axes are reversed relative to SAS and to p81
74	Figure 2.8(b): V1 distribution should be positively skewed
81	Figure 2.11: z should be \|z\| (twice)
90	Glossary entry for Reliability: remove "is intended to" and make "measure" plural
171	b_total should be b1_total
189	3rd paragraph: Table 4.9 is on page 192, not 193.
191	Exhibit1: r_y,x1(x2) eqn should have 60- in the numerator
191	Exhibit1: second r_y,x1(x2) should be r_y,x2(x1)
191	Exhibit1: r_y,x2(x1) should refer to x_2 = .11^2, not x_1
225	Line 13: +-2sqrt(n), should be +-2/sqrt(n)
229	3rd line from bottom: value should be 0.08, not 0.2
244	Expressions regarding Zjk should refer to Wij, not Wi
193	8th line: VIF value should be 10, not .10
195	Line 4: higher should be lower
199	First line in R Square section: expression should read (.7012 = .491)
201	4th line from bottom: value should be 0.014, not 0.14
241	Under "centroid" other uses of centroid (other than for Z scores) are also common. For example, the vectors of variable means for the original variables - when segregated into groups - are referred to as group centroids. The textbook prefers the name group mean as used in Table 5.1 on page 248.
241	Under "discriminant function," all W symbols should have a second subscript j for the jth discriminant function. For example, W_ij or W_ji. Also the intercepts should also be indexed with j.
245	in Figure 5.1 each figure should have the left-hand symmetrical area shaded also
290	Two corrections:
	6th line: correctly classified is 85.0 percent, not 86.4%
	Press's Q calculations: should use 55 and 34 rather than 56 and 37. This results in new Q values of 41.7 and 19.6
291	Make the following changes in Table 5.10
	The second (extra) line containing "Total Value Analysis" with values of 38, 4 and 34 should be deleted.
	The percent of grouped cases correctly classified for the holdout sample is 85%.
335	4th line above T^2 equation: "N1+N2-2-1" should be "N1+N2-p-1"

29	dummy coding for "specification buying" is unintuitively assigned
99	SAS does in fact supply partial correlation matrix
292	Add following footnote at bottom of Table 5.11 Note: Cases 82 and 64 are predicted to be in Group 0 rather than Group 1 since SPSS utilizes standardized discriminant scores when separate variance estimates are used
Chapter 5	Note on classification process: Because separate variance estimates were empolyed in the two group example, SPSS utilizes standardized Z scores in classification. This results in two cases (#82 in the analysis sample and #64 in the holdout sample) to be classified as Group 0 (Specification Buying) whereas the simple cutting score approach would classify them as Group 1 (Total Value Analysis)
	The cutting score approach described in the text is a simplified description of the classification process and may not exactly replicate actual results in the instances of separate variance estimates, use of Fishers's linear discriminant functions for classification or other options. It does, however, accurately describe the basic process and will result in identical results in most situations.