558 CHAPTER 10 NUMERICAL METHODS

C 25. (a) Compute the eigenvalues of C In Exercises 27 and 28, apply four iterations of the power method
(with scaling) to approximate the dominant eigenvalue of the given
matrix. After each iteration, scale the approximation by dividing by
3 4 3 4
2 1 2 3
A5 and B 5 . its length so that the resulting approximation will be a unit vector.
1 2 1 4
24

3 4
25. (b) Apply four iterations of the power method with scaling to 7 2
3 4
5 6
each matrix in part (a), starting with x0 5 s21, 2d. 27. A 5 28. A 5 16 29 6
4 3
25. (c) Compute the ratios l2yl1 for A and B. For which do you 8 24 5
expect faster convergence?
26. Use the proof of Theorem 10.3 to show that
AsAkx 0d < l1sAkx 0d
for large values of k. That is, show that the scale factors
obtained in the power method approach the dominant eigen-
value.

10.4 APPLICATIONS OF NUMERICAL METHODS

Applications of Gaussian Elimination with Pivoting
In Section 2.5 we used least squares regression analysis to find linear mathematical models
that best fit a set of n points in the plane. This procedure can be extended to cover poly-
nomial models of any degree as follows.

Regression Analysis The least squares regression polynomial of degree m for the points {(x1, y1), (x2, y2),
. . . , (xn, yn)} is given by
for Polynomials
y 5 amxm 1 am21xm21 1 . . . 1 a2x2 1 a1x 1 a0,
where the coefficients are determined by the following system of m 1 1 linear equa-
tions.
na0 1 sS xida1 1
sS xi2da2 1 . . . 1 sS ximdam 5 S yi
sS xida0 1 sS xi2da1 1 sS xi3da2 1 . . . 1 sS xim11dam 5 S xiyi
sS xi2da0 1 sS xi3da1 1 sS xi4da2 1 .. . . 1 sS xim12dam 5 S xi2yi
.
.
.
sS xi da0 1 sS xi da1 1 sS xi da2 1 . . . 1 sS xi dam 5 S ximyi
m m11 m12 2m

Note that if m 5 1 this system of equations reduces to

na0 1 2 sS xida1 5 S yi
sS xida0 1 sS xi2da1 5 S xiyi ,
 SECTION 10.4 APPLICATIONS OF NUMERICAL METHODS 559

which has a solution of

nS xiyi 2 (S xi)(S yid S yi Sx
a1 5 and a0 5 2 a1 i.
nS xi2 2 (S xi)2 n n
Exercise 16 asks you to show that this formula is equivalent to the matrix formula for linear
regression that was presented in Section 2.5.
Example 1 illustrates the use of regression analysis to find a second-degree polynomial
model.

EXAMPLE 1 Least Squares Regression Analysis

The world population in billions for the years between 1950 and 1985, as given by the
Statistical Abstract of the United States, is shown in the following table.

Year 1950 1955 1960 1965 1970 1975 1980 1985

Population 2.53 2.77 3.05 3.36 3.72 4.10 4.47 4.87

Find the second-degree least squares regression polynomial for these data and use the re-
sulting model to predict the world population for 1990 and 1995.

Solution We begin by letting x 5 24 represent 1950, x 5 23 represent 1955, and so on. Thus the
collection of points is given by {(24, 2.53), (23, 2.77), (22, 3.05), (21, 3.36), (0, 3.72),
(1, 4.10), (2, 4.47), (3, 4.87)}, which yields
8 8 8
n 5 8, o
i51
xi 5 24, o
i51
xi2 5 44, ox
i51
i
3
5 264,

8 8 8 8

o
i51
xi4 5 452, o
i51
yi 5 28.87, o
i51
xi yi 5 20.24, o x y 5 146.78.
i51
2
i i

Therefore the system of linear equations giving the coefficients of the quadratic model
y 5 a2x2 1 a1x 1 a0 is

88a0 2 44a1 1 444a2 5 228.87

24a0 1 44a1 2 664a2 5 20.24
44a0 2 64a1 1 452a2 5 146.78.

Gaussian elimination with pivoting on the matrix

8 24 44 28.87

3 24 44 264 20.24
44 264 452 146.78
4
560 CHAPTER 10 NUMERICAL METHODS

produces
1 21.4545 10.2727 3.3359

30 21.4545 20.6000 0.3432 .

0 20.4545 21.6000 0.0130
4
Thus by back substitution we find the solution to be
a2 5 0.0130, a1 5 0.3510, a0 5 3.7126,

Figure 10.1 and the regression quadratic is

6 1995 y 5 0.0130x2 1 0.3510x 1 3.7126.

1990
Given 5 Figure 10.1 compares this model with the given points. To predict the world population for
points Predicted
4
points 1990, we let x 5 4, obtaining
3
2 y 5 0.0130s42d 1 0.3510s4d 1 3.7126 5 5.32 billion.
1
Similarly, the prediction for 1995 sx 5 5d is
−4 −3 −2 −1 1 2 3 4 5 y 5 0.0130s52d 1 0.3510s5d 1 3.7126 5 5.79 billion.

EXAMPLE 2 Least Squares Regression Analysis

Find the third-degree least squares regression polynomial

y 5 a3x3 1 a2x2 1 a1x 1 a0
for the points
hs0, 0d, s1, 2d, s2, 3d, s3, 2d, s4, 1d, s5, 2d, s6, 4dj.

Solution For this set of points the linear system

soxna0 1 ssoxida1 1 soxi2da2 1 soxi3da3 5 oyi

ssoxida0 1 soxi2da1 1 soxi3da2 1 soxi4da3 5 oxiyi
soxi2da0 1 soxi3da1 1 soxi4da2 1 soxi5da3 5 oxi2yi
soxi3da0 1 soxi4da1 1 soxi5da2 1 soxi6da3 5 oxi3yi

becomes

427a0 1 2,221a1 1 12,291a2 1 67,441a3 5 1,214

421a0 1 2,291a1 1 12,441a2 1 62,275a3 5 1,252
491a0 1 2,441a1 1 12,275a2 1 12,201a3 5 1242
441a0 1 2,275a1 1 12,201a2 1 67,171a3 5 1,258.
 SECTION 10.4 APPLICATIONS OF NUMERICAL METHODS 561

Using Gaussian elimination with pivoting on the matrix

3 4
7 21 91 441 14
21 91 441 2,275 52
91 441 2,275 12,201 242
441 2,275 12,201 67,171 1,258

produces

3 4
1.0000 5.1587 27.6667 152.3150 2.8526
0.0000 1.0000 8.5312 58.3482 0.6183
,
Figure 10.2 0.0000 0.0000 1.0000 9.7714 0.1286
y 0.0000 0.0000 0.0000 1.0000 0.1667
(6, 4)
4 which implies that
(2, 3) (5, 2)
3
(3, 2) a3 5 0.1667, a2 5 21.5000, a1 5 3.6905, a0 5 20.0714.
2
(1, 2) Therefore the cubic model is
1
(0, 0) (4, 1) y 5 0.1667x3 2 1.5000x2 1 3.6905x 2 0.0714.
x
1 2 3 4 5 6 Figure 10.2 compares this model with the given points.

Applications of the Gauss-Seidel Method

EXAMPLE 3 An Application to Probability

Figure 10.3 is a diagram of a maze used in a laboratory experiment. The experiment is

begun by placing a mouse at one of the ten interior intersections of the maze. Once the
mouse emerges in the outer corridor, it cannot return to the maze. When the mouse is at an
interior intersection, its choice of paths is assumed to be random. What is the probability
that the mouse will emerge in the “food corridor” when it begins at the ith intersection?

Solution We let the probability of winning (getting food) by starting at the ith intersection be repre-
sented by pi. Then we form a linear equation involving pi and the probabilities associated
Figure 10.3
with the intersections bordering the ith intersection. For instance, at the first intersection the
mouse has a probability of 41 of choosing the upper right path and losing, a probability of 41
of choosing the upper left path and losing, a probability of 41 of choosing the lower right
1
path (at which point it has a probability of p3 of winning), and a probability of 14 of choos-
2 3 ing the lower left path (at which point it has a probability of p2 of winning). Thus we have
4 5 6 p1 5 14 s0d 1 14 s0d 1 14 p2 1 14 p3.
{
{

{
{

7 8 9 10
Upper Upper Lower Lower
right left left right
Food
562 CHAPTER 10 NUMERICAL METHODS

Using similar reasoning, we find the other nine probabilities to be represented by the fol-
lowing equations.
p2 5 15 s0d 1 15 p1d 1 15 p3 1 15 p4 1 15 p5
p3 5 15 s0d 1 15 p1d 1 15 p2 1 15 p5 1 15 p6
p4 5 15 s0d 1 15 p2d 1 15 p5 1 15 p7 1 15 p8
p5 5 16 p2d 1 16 p3d 1 16 p4 1 16 p6 1 16 p81 1 16 p9
p6 5 15 s0d 1 15 p3d 1 15 p5 1 15 p9 1 15 p10
p7 5 14 s0d 1 14 s1d 1 14 p4 1 14 p8
p8 5 15 s1d 1 15 p4d 1 15 p5 1 15 p7 1 15 p9
p9 5 15 s1d 1 15 p5d 1 15 p6 1 15 p8 1 15 p10
p10 5 14 s0d 1 14 s1d 1 14 p6 1 14 p9

Rewriting these equations in standard form produces the following system of ten linear
equations in ten variables.
4p1 2 5p2 2 5p3 2 5p4 2 6p5 2 5p6 2 4p7 2 5p8 2 5p9 2 4p10 5 0
2p1 1 5p2 2 5p3 2 5p4 2 6p5 2 5p6 2 5p7 2 5p8 2 5p9 2 4p10 5 0
2p1 2 5p2 1 5p3 2 5p4 2 5p5 2 5p6 2 5p7 2 5p8 2 5p9 2 5p10 5 0
4p1 22p2 1 5p3 1 5p4 2 5p5 2 5p6 2 5p7 2 5p8 2 5p9 2 5p10 5 0
4p1 22p2 2 5p3 2 5p4 1 6p5 2 6p6 2 5p7 2 5p8 2 5p9 2 5p10 5 0
4p1 2 4p2 22p3 2 4p4 2 5p5 1 5p6 2 5p7 2 5p8 2 5p9 2 5p10 5 0
4p1 2 5p2 2 5p3 22p4 2 6p5 2 5p6 1 4p7 2 5p8 2 5p9 2 4p10 5 1
4p1 2 5p2 2 5p3 22p4 2 6p5 2 5p6 2 4p7 1 5p8 2 5p9 2 4p10 5 1
4p1 2 5p2 2 5p3 2 2p4 22p5 2 5p6 2 4p7 2 5p8 1 5p9 2 4p10 5 1
4p1 2 5p2 2 5p3 2 2p4 2 2p5 22p6 2 4p7 2 5p8 2 5p9 1 4p10 5 1

The augmented matrix for this system is

4 21 21 0 0 0 0 0 0 0 0
21 5 21 21 21 0 0 0 0 0 0
21 21 5 0 21 21 0 0 0 0 0
0 21 0 5 21 0 21 21 0 0 0
0 21 21 21 6 21 0 21 21 0 0 .
0 0 21 0 21 5 0 0 21 21 0
0 0 0 21 0 0 4 21 0 0 1
0 0 0 21 21 0 21 5 21 0 1
0 0 0 0 21 21 0 21 5 21 1
0 0 0 0 0 21 0 0 21 4 1
 SECTION 10.4 APPLICATIONS OF NUMERICAL METHODS 563

Using the Gauss-Seidel method with an initial approximation of p1 5 p2 5 . . . 5 p10 5 0

produces (after 18 iterations) an approximation of
p1 5 0.090, p2 5 0.180
p3 5 0.180, p4 5 0.298
p5 5 0.333, p6 5 0.298
p7 5 0.455, p8 5 0.522
p9 5 0.522, p10 5 0.455.

The structure of the probability problem described in Example 3 is related to a technique

called finite element analysis, which is used in many engineering problems.
Note that the matrix developed in Example 3 has mostly zero entries. We call such ma-
trices sparse. For solving systems of equations with sparse coefficient matrices, the Jacobi
and Gauss-Seidel methods are much more efficient than Gaussian elimination.

Applications of the Power Method

Section 7.4 introduced the idea of an age transition matrix as a model for population
growth. Recall that this model was developed by grouping the population into n age classes
of equal duration. Thus for a maximum life span of L years, the age classes are given by
the following intervals.
First age Second age nth age
class class class
{
{

{
(n 2 l)L
30, n2, 3n, n 2, . . . , 3 4
L L 2L
,L
n
The number of population members in each age class is then represented by the age distri-
bution vector

34
x1 Number in first age class
x Number in second age class
x 5 .2 .
..
xn Number in nth age class

Over a period of Lyn years, the probability that a member of the ith age class will survive
to become a member of the si 1 1dth age class is given by pi, where 0 # pi # 1, i 5 1, 2,
. . . , n 2 1. The average number of offspring produced by a member of the ith age class
is given by bi, where 0 # bi, i 5 1, 2, . . . , n. These numbers can be written in matrix form:

b1 b2 b3 . . . bn21 bn
p1 0 0 ... 0 0
A 5 0 p2 0 ... 0 0
.. .. .. .. ..
. . . . .
0 0 0 . . . pn21 0
564 CHAPTER 10 NUMERICAL METHODS

Multiplying this age transition matrix by the age distribution vector for a given period of
time produces the age distribution vector for the next period of time. That is,
Axi 5 xi11.
In Section 7.4 we saw that the growth pattern for a population is stable if the same per-
centage of the total population is in each age class each year. That is,
Axi 5 xi11 5 lxi.
For populations with many age classes, the solution to this eigenvalue problem can be
found with the power method, as illustrated in Example 4.

EXAMPLE 4 A Population Growth Model

Assume that a population of human females has the following characteristics.

Average Number of Probability of
Age Class Female Children Surviving to
(in years) During Ten-Year Period Next Age Class

00 # age , 110 0.000 0.985

10 # age , 120 0.174 0.996
20 # age , 130 0.782 0.994
30 # age , 140 0.263 0.990
40 # age , 150 0.022 0.975
50 # age , 160 0.000 0.940
60 # age , 170 0.000 0.866
70 # age , 180 0.000 0.680
80 # age , 190 0.000 0.361
90 # age , 100 0.000 0.000
Find a stable age distribution for this population.

Solution The age transition matrix for this population is

0.000 0.174 0.782 0.263 0.022 0.000 0.000 0.000 0.000 0.000
0.985 0 0 0 0 0 0 0 0 0
0 0.996 0 0 0 0 0 0 0 0
0 0 0.994 0 0 0 0 0 0 0
A5 0 0 0 0.990 0 0 0 0 0 0 .
0 0 0 0 0.975 0 0 0 0 0
0 0 0 0 0 0.940 0 0 0 0
0 0 0 0 0 0 0.866 0 0 0
0 0 0 0 0 0 0 0.680 0 0
0 0 0 0 0 0 0 0 0.361 0
 SECTION 10.4 APPLICATIONS OF NUMERICAL METHODS 565

To apply the power method with scaling to find an eigenvector for this matrix, we use an
initial approximation of x05(1, 1, 1, 1, 1, 1, 1, 1, 1, 1). The following is an approximation
for an eigenvector of A, with the percentage of each age in the total population.
Percentage in
Eigenvector Age Class Age Class

1.000 00 # age , 110 15.27

0.925 10 # age , 120 14.13
0.864 20 # age , 130 13.20
0.806 30 # age , 140 12.31
x5 0.749 40 # age , 150 11.44
0.686 50 # age , 160 10.48
0.605 60 # age , 170 19.24
0.492 70 # age , 180 17.51
0.314 80 # age , 190 14.80
0.106 90 # age , 100 11.62

The eigenvalue corresponding to the eigenvector x in Example 4 is l ø 1.065. That is,

1.000 1.065 1.000

0.925 0.985 0.925
0.864 0.921 0.864
0.806 0.859 0.806
0.749 0.798 0.749
Ax 5 A ø ø 1.065 .
0.686 0.731 0.686
0.605 0.645 0.605
0.492 0.524 0.492
0.314 0.334 0.314
0.106 0.113 0.106

This means that the population in Example 4 increases by 6.5% every ten years.

REMARK: Should you try duplicating the results of Example 4, you would notice that
the convergence of the power method for this problem is very slow. The reason is that the
dominant eigenvalue of l ø 1.065 is only slightly larger in absolute value than the next
largest eigenvalue.
 566 CHAPTER 10 NUMERICAL METHODS

SECTION 10.4 ❑ EXERCISES

Applications of Gaussian Elimination with Pivoting (a) Find the least squares regression line for these data.
C In Exercises 1–4, find the second-degree least squares regression (b) Find the second-degree least squares regression polyno-
mial for these data.
polynomial for the given data. Then graphically compare the model
to the given points. (c) Sketch the graphs of the models found in parts (a) and (b).
(d) Use the models found in parts (a) and (b) to approximate
1. s22, 1d, s21, 0d, s0, 0d, s1, 1d, s3, 2d
the maximum number of minutes a diver should stay at a
2. s0, 4d, s1, 2d, s2, 21d, s3, 0d, s4, 1d, s5, 4d depth of 120 feet. ( The value given in the Navy’s tables
3. s22, 1d, s21, 2d, s0, 6d, s1, 3d, s2, 0d, s3, 21d is 15 minutes.)
4. s1, 1d, s2, 1d, s3, 0d, s4, 21d, s5, 24d C 12. The life expectancy for additional years of life for females in
C In Exercises 5–8, find the third-degree least squares regression poly- the United States is given in the following table.
nomial for the given data. Then graphically compare the model to the Current Age 10 20 30 40
given points.
Life Expectancy 69.7 59.9 50.2 40.6
5. (0, 0), (1, 2), (2, 4), (3, 1), (4, 0), (5, 1)
6. (1, 1), (2, 4), (3, 4), (5, 1), (6, 2)
7. s23, 4d, s21, 1d, s0, 0d, s1, 2d, s2, 5d Current Age 50 60 70 80
8. s27, 2d, s23, 0d, s1, 21d, s2, 3d, s4, 6d Life Expectancy 31.3 22.8 15.3 9.0
C 9. Find the second-degree least squares regression polynomial for
the points Find the second-degree least squares regression polynomial
for these data.
p p 1 p 1 p
12 2 , 02, 12 3 , 22, s0, 1d, 1 3 , 22, 1 2 , 02. C 13. Total federal payments (in billions of dollars) for Medicare
from 1970 to 1984 are given in the following table.
Then use the results to approximate cos spy4d. Compare the
approximation with the exact value. Year 1970 1972 1974 1976
C 10. Find the third-degree least squares regression polynomial for Medicare 7.5 9.1 13.1 19.3
the points
p p p p
12 4 , 212, 12 3 , 2Ï32, s0, 0d, 1 3 , Ï32, 1 4 , 12.
Year 1978 1980 1982 1984

Medicare 25.9 36.8 52.4 64.8

Then use the result to approximate tan spy6d. Compare the
approximation with the exact value. Find the second-degree least squares regression polynomial
C 11. The number of minutes a scuba diver can stay at a particular for these data. Then use the result to predict the Medicare
depth without acquiring decompression sickness, as given by payments for 1986 and 1988.
the United States Navy’s Standard Air Decompression Tables, C 14. Rework Exercise 13 using the data in the following table, in
is shown in the following table. which Medicare payments have been adjusted for inflation.

Depth (in feet) 35 40 50 60 70 Year 1970 1972 1974 1976

Time (in minutes) 310 200 100 60 50 Medicare 7.5 8.4 10.3 13.2

Depth (in feet) 80 90 100 110 Year 1978 1980 1982 1984

Time (in minutes) 40 30 25 20 Medicare 15.4 17.3 21.1 24.2

 SECTION 10.4 EXERCISES 567

C 15. Find the least squares regression line for the population data C 19. A square metal plate has a constant temperature on each of its
given in Example 1. Then use the model to predict the world four boundaries, as shown in Figure 10.6. Use a 4 3 4 grid to
population in 1990 and 1995, and compare the results with approximate the temperature distribution in the interior of the
the predictions obtained in Example 1. plate. Assume that the temperature at each interior point is the
16. Show that the formula for the least squares regression line average of the temperatures at the four closest neighboring
presented in Section 2.5 is equivalent to the formula presented points.
in this section. That is, if Figure 10.6

34 3 4
y1 1 x1 100°

3 4
y 1 x2 a
Y 5 .2 , X 5 .. , A5 0 ,
.. . ... a1 1 2 3
yn 1 xn
100° 4 5 6 0°
then the matrix equation A 5 sXTXd21XTY is equivalent to
nS xiyi 2 (S xi)(S yi) S yi Sx 7 8 9
a1 5 and a0 5 2 a1 i.
nS xi2 2 (S xi)2 n n
0°
Applications of the Gauss-Seidel Method
C 17. Suppose that the experiment in Example 3 is performed with
C 20. A rectangular metal plate has a constant temperature on each
the maze shown in Figure 10.4. Find the probability that the
mouse will emerge in the food corridor when it begins in the of its four boundaries, as shown in Figure 10.7. Use a 4 3 5
ith intersection. grid to approximate the temperature distribution in the interior
Figure 10.4 of the plate. Assume that the temperature at each interior point
is the average of the temperatures at the four closest neighbor-
ing points.
Figure 10.7
1 2 3 120°

1 2 3 4
4 5 6
80° 5 6 7 8 40°
Food 9 10 11 12

C 18. Suppose that the experiment in Example 3 is performed with

the maze shown in Figure 10.5. Find the probability that the 0°
mouse will emerge in the food corridor when it begins in the
ith intersection. Applications of the Power Method
Figure 10.5
In Exercises 21–24, the given matrix represents the age transition
matrix for a population. Use the power method with scaling to find
a stable age distribution.
1 2 3
3 4 3 4
1 4 1 2
21. A 5 1 22. A 5 1
2 0 4 0
4 5 6

3 4 3 4
0 1 2 1 2 2
C 23. A 5 1
0 0 C 24. A 5 1
0 0
2 3
7 8 9 0 1
0 0 1
0
4 3

Food
 568 CHAPTER 10 NUMERICAL METHODS

C 25. In Example 1 of Section 7.4, a laboratory population of rab- 28. In Example 2 of Section 2.5, the stochastic matrix
bits is described. The age transition matrix for the population is

3 4
0.70 0.15 0.15

3 4
0 6 8 P 5 0.20 0.80 0.15
A 5 0.5 0 0 . 0.10 0.05 0.70
0 0.5 0
represents the transition probabilities for a consumer prefer-
Find a stable age distribution for this population. ence model. Use the power method to approximate a domi-
C 26. A population has the following characteristics. nant eigenvector for this matrix. How does the approximation
(a) A total of 75% of the population survives its first year. Of relate to the steady-state matrix described in the discussion
that 75%, 25% survives its second year. The maximum following Example 3 in Section 2.5?
life span is three years. C 29. In Exercise 9 of Section 2.5, a population of 10,000 is
(b) The average number of offspring for each member of the divided into nonsmokers, moderate smokers, and heavy
population is 2 the first year, 4 the second year, and 2 the smokers. Use the power method to approximate a dominant
third year. eigenvector for this matrix.
Find a stable age distribution for this population. (See
Exercise 9, Section 7.4.)
27. Apply the power method to the matrix

31 4
1 1
A5
0
discussed in Chapter 7 (Fibonacci sequence). Use the power
method to approximate the dominant eigenvalue of A. (The
dominant eigenvalue is l 5 s1 1 Ï5dy2.)