Answers to Problems

1.

a.

To estimate the costs, calculate the number of square feet in the

exterior of the house and enter that amount into the equation.
Calculate the number of square feet in the exterior of the house:
2(9x40)+ 2(9x28) = 1,224
Estimated Cost = $15 + ($0.06/s.f.)(1,224 s.f.)
= $88.44

b.

If the cost per square increases, the new cost should be used to
estimate costs.
Estimated Cost = $15 + ($0.064/s.f.)(1,224 s.f.)
= $93.34

c.

The amount of time it takes to do the job could be affected by the

type of siding on the house and whether the house is one or two
stories. The type of siding might also affect the amount of
cleaning supplies needed. Another factor that could affect costs
is the distance that a crew must travel to get to a job.

3.
PERIOD
2003
2004
2005
2006
2007
2008
2009
2010
2011

CODE (X)
1
2
3
4
5
6
7
8
9
-45

SUM

b =

BEEF IMPORTED(Y)
82
101
114
126
137
151
164
182
189
----1,246

n(X)(Y) - X Y
--------------------nX2 - (X)2
(9)(7,026) - (45)(1,246)
-------------------------(9)(285) - (45)2

= 13.27 thousands of tons of beef/year

a =

Y
---n
1,246
-------

b(X)
------n
13.27(45)
------------

(X)(Y)
82
202
342
504
685
906
1,148
1,456
1,701
----7,026

X2
1
4
9
16
25
36
49
64
81
--285

9
=

72.1 thousands of tons of beef

Imports of beef are projected to increase 13.27 thousands of tons per

year. The interception is 72.1 thousands, and it represents the point on
the regression line for 1992.
Y e = 72.1 + 13.27(X)
The periods are consecutively numbered, 2012 is coded as 10, and thus,
the projected amount of beef imported for year 2012 is:
Y 12 = 72.1 + (13.27)(10)
= 204.8 thousands of tons
5.
a.
To calculate the MSE and the MAD, the forecasted
amounts for December through July must be calculated. The parameters for this
model are calculated in the previous problem.
The periods are consecutively numbered, December is coded as 1,
so, the projected number of hours for December is:

PERIOD
December
January
February
March
April
May
June
July
SUM

Y De

= 291.1 + (158.1)(1)
= 449.2 hours

Y Ja

= 291.1 + (158.1)(2)
= 607.3 hours

Y Fe

= 291.1 + (158.1)(3)
= 765.4 hours

Y Mr

= 291.1 + (158.1)(4)
= 923.5 hours

Y Ap

= 291.1 + (158.1)(5)
= 1,081.6 hours

Y My

= 291.1 + (158.1)(6)
= 1,239.7 hours

Y Ju

= 291.1 + (158.1)(7)
= 1,397.8 hours

Y Jy

= 291.1 + (158.1)(8)
= 1,555.9 hours

ACTUAL HRS
OF SERVICE
300
750
650
920
1,300
1,400
1,200
1,500

FORECASTED HRS
OF SERVICE
449.2
607.3
765.4
923.5
1,081.6
1,239.7
1,397.8
1,555.9

ERROR
-149.2
142.7
-115.4
- 3.5
218.4
160.3
-197.8
- 55.9
------.4

SQUARED
ERROR
22,260.64
20,363.29
13,317.16
12.25
47,698.56
25,696.09
39,124.84
3,124.81
----------171,597.64

ABSOLUTE
ERROR
149.2
142.7
115.4
3.5
218.4
160.3
197.8
55.9
------1,043.2

MSE

n
(x t - f t )2
t=1
= -----------------n
171,597.64
= -----------------8
= 21,449.7

MAD

n
|x t - f t |
t=1
= -----------------n
1,043.2
= -----------------8
= 130.4

7.

b.

The model provides reasonably good predictions based on the MSE

and MAD. The MAD is easier to interpret because it is an estimate
of the average error in the forecast.

a.

When A = .1
f t = A (x t-1 ) + (1-A) f t-1
For the initial value, we assume
f 01
f 02
f 03
f 04
f 05
f 06
f 07
f 08
f 09
f 10
f 11
f 12

b.

=
=
=
=
=
=
=
=
=
=
=
=

46.0 mergers
.1(46) + .9(46.0)
.1(62) + .9(46.0)
.1(45) + .9(47.6)
.1(64) + .9(47.3)
.1(61) + .9(49.0)
.1(83) + .9(50.2)
.1(123) + .9(53.5)
.1(97) + .9(60.4)
.1(186) + .9(64.1)
.1(225) + .9(76.3)
.1(240) + .9(91.2)

=
=
=
=
=
=
=
=
=
=
=

f 01 = x 00

46.0
47.6
47.3
49.0
50.2
53.5
60.4
64.1
76.3
91.2
106.0

When A = .7
f 01
f 02
f 03
f 04
f 05

=
=
=
=
=

46.0 mergers
.7(46)
+ .3(46.0)
.7(62)
+ .3(46.0)
.7(45)
+ .3(57.2)
.7(64)
+ .3(48.7)

=
=
=
=

46.0
57.2
48.7
59.4

f 06
f 07
f 08
f 09
f 10
f 11
f 12

=
=
=
=
=
=
=

.7(61)
.7(83)
.7(123)
.7(97)
.7(186)
.7(225)
.7(240)

+
+
+
+
+
+
+

.3(59.4)
= 60.5
.3(60.5)
= 76.3
.3(76.3) = 109.0
.3(109.0) = 100.6
.3(100.6) = 160.0
.3(160.4) = 205.6
.3(205.6) = 229.7

The graph shows that the high smoothing constant, A = .7, reacts more quickly
to fluctuations in demand. The low value of A causes the forecast to
react very slowly. Thus, the forecast is smoother.

9.

a.

First, convert the time variable to a simpler number. The number

of appliances returned (Y) depends on time (X), the independent
variable.

PERIOD

CODE(X)

5 yr,q1
q2
q3
q4

1
2
3
4

APPLIANCES RETURN(Y)
1.2
0.8
0.6
1.1

(X)(Y)

X2

1.2
1.6
1.8
4.4

1
4
9
16

4 yr,q1
q2
q3
q4
3 yr,q1
q2
q3
q4
2 yr,q1
q2
q3
q4
1 yr,q1
q2
q3
q4
SUM

5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
--210

1.7
1.2
1.0
1.5
3.1
3.5
3.5
3.2
2.6
2.2
1.9
2.5
2.9
2.5
2.2
3.0
---42.2

8.5
7.2
7.0
12.0
27.9
35.0
38.5
38.4
33.8
30.8
28.5
40.0
49.3
45.0
41.8
60.0
-----512.7

25
36
49
64
81
100
121
144
169
196
225
256
289
324
361
400
----2,870

n(X)(Y) - X Y
b = -------------------nX2 - (X)2
(20)(512.7) - (210)(42.2)
= -------------------------(20)(2,870) - (210)2
1,392
= -----13,300
= 0.1047 thousands of appliances/quarter

Y
b(X)
a = ----- - --------n
n
42.2
(.1047)(210)
= ------ - -----------20
20
= 1.011 thousands of appliances
As a result, the number of appliances returned is projected to
increase 0.1047 thousand per quarter. The y-intercept is 1.011
thousand, and it represents the point on the regression line for
the quarter prior to the first quarter for which data are
available. The model is:
Y e = 1.011 + .1047(X)
b.

Because the periods are consecutively numbered, the second quarter

of the current year is coded as 22; the projected number of
appliances returned for the second quarter of the current year is:
Y 22 = 1.011 + .1047(22)
= 3.314 thousand

11.

f t = A (x t-1 ) + (1-A) f t-1

f Oct10
f Nov10
f Dec10
f Jan11
f Feb11
f Mar11
f Apr11
f May11
f Jun11
f Jul11
f Aug11
f Sep11
f Oct11
f Nov11
f Dec11
f Jan12

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

55.0 thousands of
.8(53) + .2(55.0)
.8(60) + .2(53.4)
.8(49) + .2(58.7)
.8(48) + .2(50.9)
.8(61) + .2(48.6)
.8(61) + .2(58.5)
.8(53) + .2(60.5)
.8(63) + .2(54.5)
.8(53) + .2(61.3)
.8(51) + .2(54.7)
.8(60) + .2(51.7)
.8(58) + .2(58.4)
.8(52) + .2(58.1)
.8(51) + .2(53.2)
.8(63) + .2(51.4)

units
= 53.4
= 58.7
= 50.9
= 48.6
= 58.5
= 60.5
= 54.5
= 61.3
= 54.7
= 51.7
= 58.3
= 58.1
= 53.2
= 51.4
= 60.7

The forecast depends too much on the most recent months' actual demand
because the smoothing constant (A) is too large. This causes the
forecast to change significantly in response to any change in the prior
periods actual demand.

13.
PERIOD
2010, 9
10
11
12
2011, 1
2
3
4
5
6
7
8
9
10
11
12
SUM

CODE(X)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
--136

UNIT SOLD(Y)
55
53
60
49
48
61
61
53
63
53
51
60
58
52
51
63
---891

n(X)(Y) - (X)(Y)
b = ----------------------

(X)(Y)
55
106
180
196
240
366
427
424
567
530
561
720
754
728
765
1,008
------7,627

X2
1
4
9
16
25
36
49
64
81
100
121
144
169
196
225
256
------1,496

nX2 - (X)2
(16)(7,627) - (136)(891)
= -------------------------(16)(1,496) - (136)2
856
= -----5,440
= 0.157 thousands of units/month

Y
a = ---n

b(X)
------n

891
= ---16

(0.157)(136)
------------16

= 54.4 thousands of units

Therefore the regression line is:
Y e = 54.4 + 0.157(X)
The periods are consecutively numbered, so Jan. 2012 is coded as 17,
Feb. 2012 is coded as 18,....., June 2012 is coded as 22.
f Jan12
f Feb12
f Mar12
f Apr12
f May12
f Jun12

15.

=
=
=
=
=
=

54.4
54.4
54.4
54.4
54.4
54.4

+
+
+
+
+
+

0.157(17)
0.157(18)
0.157(19)
0.157(20)
0.157(21)
0.157(22)

=
=
=
=
=
=

57.07 thousands of units

57.23
57.38
57.54
57.70
57.85

a.
Observation
1
2
3
4
5
6
7
8
9
10
11
12
13

X
154
265
540
332
551
487
305
218
144
155
242
234
343
-----

Y
743
830
984
801
964
955
839
478
720
782
853
878
940
------

(X)(Y)
114,422
219,950
531,360
265,932
531,164
465,085
255,895
104,204
103,680
121,210
206,426
205,452
322,420
---------

X2
23,716
70,225
291,600
110,224
303,601
237,169
93,025
47,524
20,736
24,025
58,564
54,756
117,649
---------

Y2
552,049
688,900
968,256
641,601
929,296
912,025
703,921
228,484
518,400
611,524
727,609
770,884
883,600
---------

SUM

3,970

10,767

3,447,200

1,452,814

9,136,549

n(X)(Y) - (X)(Y)
b = ---------------------nX2 - (X)2
(13)(3,447,200) - (3,970)(10,767)
b = --------------------------------(13)(1,452,814) - (3,970)2
2,068,610
= ---------3,125,682
= .6618
Y
b(X)
a = -- - ----n
n
10,767
a = -----13

(.662)(3,970)
------------13

= 626.1
Y e = 626.1 + .6618(X)

b.

r=

r=

n[ (X)(Y)] - ( X)( Y)
[n( X 2 ) - ( X)2 ][n( Y 2 ) - ( Y)2 ]

(13)(3,447,200) - (3,970)(10,767)
[(13)(1,452,814) - (3,970 )2 ][(13)(9,136,549) - (10,767)2 ]

2,068,610
r = ---------------2,983,008.8
=

0.693

s=

Y 2 - a Y - b (X)(Y)
n-2

s=

(9,136,549) - (626.1)(10,767) - (.6618)(3,447,200)

11
= 101.8

c.

The standard error of the estimate is a measure of the amount of

dispersion around the regression line. The correlation coefficient
measures the strength of the relationship between the variables.
The higher the correlation coefficient (r) the stronger the
relationship. See response to 14b.