Forecasting l CHAPTER 5

TRUE/FALSE
5.1

Time-series models rely on judgment in an attempt to incorporate qualitative or subjective factors

into the forecasting model.
ANSWER: FALSE

*5.2

Use of any forecasting procedure is somewhat subjective.

ANSWER: TRUE

5.3

The coefficient of correlation expresses the degree or strength of a linear relationship.

ANSWER: TRUE

*5.4

To make a forecast which is accurate over time requires that we collect data over time.
ANSWER: TRUE

5.5

One of the most popular qualitative forecasting methods is the Delphi technique.
ANSWER: TRUE

5.6

A disadvantage of the Delphi technique is that results are obtained slowly.

ANSWER: TRUE

5.7

Often, a variety of dependent variables may be successfully used in a linear regression forecast of a
single independent variable.
ANSWER: FALSE

5.8

A moving average forecasting method is a causal forecasting method.

ANSWER: FALSE

5.9

An exponential forecasting method is a time-series forecasting method.

ANSWER: TRUE

5.10

A trend projection forecasting method is a causal forecasting method.

ANSWER: FALSE

5.11

Tupperware International has successfully identified a single forecasting tool to predict their
companys product sales.
ANSWER: FALSE

5.12

A scatter diagram is useful to determine if a relationship exists between two variables.

ANSWER: TRUE

118

Forecasting l CHAPTER 5

5.13

A seasonal index must be between 1 and +1.

ANSWER: FALSE

5.14

Time-series models enable the forecaster to include specific representations of various qualitative
and quantitative factors.
ANSWER: FALSE

5.15

Qualitative models produce forecasts that are little better than simple guesses or coin tosses.
ANSWER: FALSE

5.16

If you need to develop a forecast in a hurry, you probably should not contemplate using the Delphi
method.
ANSWER: TRUE

5.17

If you need to develop a forecast of sales as a function of advertising expenditure and product
selling price, you should probably consider using one of the regression analysis models.
ANSWER: TRUE

5.18

One of the benefits of the Delphi method is that no one forecaster is able to unduly influence any
other forecaster.
ANSWER: TRUE

5.19

When one plots a scatter diagram, the independent variable (X) is always time.
ANSWER: FALSE

5.20

One of the benefits of using a causal forecasting model is that we are able to eliminate the impact
of random error.
ANSWER: FALSE

5.21

The fewer the periods over which one takes a moving average, the more accurately the resulting
forecast mirrors the actual data.
ANSWER: TRUE

5.22

5.23

An advantage of exponential smoothing over a simple moving average is that exponential

smoothing requires one to retain less data.
ANSWER: TRUE
An advantage of exponential smoothing over a simple moving average is that the exponential
smoothing model can be extended to include a trend term.
ANSWER: TRUE

5.24

The notion of a seasonal index can only be associated with time-series forecasting.

119

Forecasting l CHAPTER 5

ANSWER: FALSE
5.25

A correlation coefficient of +0.75 implies that the forecasted variable increases as the independent
variable increases.
ANSWER: TRUE

5.26

The purpose of a tracking signal is to help us estimate the forecast error at each data point.
ANSWER: FALSE

5.27

Adaptive smoothing is analogous to exponential smoothing where the coefficients and are
periodically updated to improve the forecast.
ANSWER: TRUE

*5.28

One of the advantages of using a scatter diagram is that it may suggest types of formatting
techniques that are appropriate.
ANSWER: TRUE

*5.29

One of the advantages of using a scatter diagram is that it may suggest types of formatting
techniques that are not appropriate.
ANSWER: TRUE

*5.30

As a causal method, moving averages are preferable to exponential smoothing.

ANSWER: FALSE

MULTIPLE CHOICE
5.31

A weighted moving average having the early periods more heavily weighted
(a)
(b)
(c)
(d)
(e)

is more responsive to recent demand changes.

is less responsive to recent demand.
places emphasis on the past demand data.
is always the most effective weighting scheme.
is not a time-series model.

ANSWER: b
5.32

One statistical method for developing a linear trend line is

(a)
(b)
(c)
(d)
(e)

"eyeballing."
the exponential smoothing method.
the causal forecasting method.
the MAD technique.
least squares.

ANSWER: e

120

Forecasting l CHAPTER 5

5.33

In the linear regression equation, "b" is the

(a)
(b)
(c)
(d)
(e)

smoothing constant.
Y-axis intercept.
slope of the regression line.
independent variable.
dependent variable.

ANSWER: c
5.34

Which of the following is not classified as a qualitative forecasting model?

(a)
(b)
(c)
(d)
(e)

exponential smoothing
Delphi
executive opinion
sales force composite
consumer market survey

ANSWER: a
5.35

Which of the following is (are) not characteristic of the scatter diagram?

(a)
(b)
(c)
(d)
(e)

The independent variable is usually measured on the horizontal (X) axis.

The dependent variable is usually measured on the vertical (y) axis.
It is useful to get a quick idea as to whether any relationship exists.
It is helpful in determining what is cause and what is effect.
none of the above

ANSWER: d
5.36

Which of the following university/commercial statistical computer packages has a forecasting

technique?
(a)
(b)
(c)
(d)
(e)

5.37

BIOMED
SAS
SPSS
Minitab
all of the above

ANSWER: e
One thing not true about the coefficient of correlation is that it
(a)
(b)
(c)
(d)
(e)

can be squared to get the coefficient of determination.

can be any number between -1 and +1.
expresses the degree or strength of the relationship between variables.
can be used to calculate next year's sales.
is usually expressed as r.

ANSWER: d
5.38

If computing a causal linear regression model of Y = a + bX and the resultant r 2 is very near zero,
then one would be able to conclude that

121

Forecasting l CHAPTER 5

(a)
(b)
(c)
(d)
(e)

Y = a + bX is a good forecasting method.

Y = a + bX is not a good forecasting method.
a multiple linear regression model is a good forecasting method for the data.
a multiple linear regression model is not a good forecasting method for the data.
none of the above

ANSWER: b
5.39

Daily demand for newspapers for the last 10 days has been as follows: 12, 13, 16, 15, 12, 18, 14,
12, 13, 15. Forecast sales for the next day using a 2-day moving average.
(a)
(b)
(c)
(d)
(e)

14
13
15
28
none of the above

ANSWER: a
5.40

Daily demand for newspapers for the last 10 days has been as follows: 12, 13, 16, 15, 12, 18, 14,
12, 13, 15. Forecast sales for the next day using a 3-day weighted moving average where the
weights are 3, 1, and 1 (the highest weight is for the most recent number).
(a)
(b)
(c)
(d)
(e)

12.8
13.0
70.0
14.0
none of the above

ANSWER: d

122

Forecasting l CHAPTER 5

5.41

Enrollment in a particular class for the last four semesters had been 120, 126, 110, and 130.
Develop a forecast of enrollment next semester using exponential smoothing with an alpha = 0.2.
Assume that an initial forecast for the first semester was 120 (so the forecast and the actual were
the same).
(a)
(b)
(c)
(d)
(e)

118.96
121.17
130
120
none of the above

ANSWER: b
5.42

Enrollment in a particular class for the last four semesters had been 120, 126, 110, and 130.
Suppose a 1-semester moving average was used to forecast enrollment (this is sometimes referred
to as a naive forecast). Thus, the forecast for the second semester would be 120, for the third
semester it would be 126, and for the last semester it would be 110. What would the MSE be for
this situation?
(a)
(b)
(c)
(d)
(e)

196.00
230.67
100.00
42.00
none of the above

ANSWER: b
5.43

A tracking signal was calculated for a particular set of demand forecasts. This tracking signal was
positive. This would indicate that
(a)
(b)
(c)
(d)
(e)

demand is greater than the forecast.

demand is less than the forecast.
demand is equal to the forecast.
the MAD is negative.
none of the above

ANSWER: a
5.44

Regression was used to develop a model to predict sales based on advertising dollars spent. The
equation developed is Y = 1000 + 20X, where X is advertising dollars and Y is sales. If $300 is
spent on advertising, what would be the best prediction for sales?
(a)
(b)
(c)
(d)
(e)

$1,600
$7,000
$1,620
$6,000
none of the above

ANSWER: b
5.45

Regression was used to develop a model to predict sales based on advertising dollars spent. The
equation developed is Y = 1000 + 20X - 2Z, where X is advertising dollars spent by your company,

123

Forecasting l CHAPTER 5

Z is the price for the product, and Y is sales. If $800 is spent by your company on advertising, and
the price is set at $100, what would be the best prediction for sales?
(a)
(b)
(c)
(d)
(e)

$17,200
$6,800
$7,200
$16,800
none of the above

ANSWER: d
5.46

An exponential smoothing model having a large

(a)
(b)
(c)
(d)
(e)

is more responsive to recent demand changes.

is less responsive to recent demand.
places emphasis on the past demand data.
is always the most effective weighting scheme.
is not a time-series model.

ANSWER: a
5.47

In the exponential smoothing forecasting method, is the

(a)
(b)
(c)
(d)
(e)

slope of the trend line.

new forecast.
Y-axis intercept.
independent variable.
trend smoothing constant.

ANSWER: e
5.48

Which of the following is a technique used to determine forecasting accuracy?

(a)
(b)
(c)
(d)
(e)

Mean deviation
Squared Average Error
Mean Absolute Percent Error
Delphi Method
none of the above

ANSWER: c

124

Forecasting l CHAPTER 5

5.49

Calculation of a correlation coefficient is part of ________ analysis.

(a)
(b)
(c)
(d)
(e)

exponential smoothing
time-series
seasonal
marginal
regression

ANSWER: e
5.50

Which of the following is not a characteristic of regression analysis?

(a) The independent variable is usually called X.
(b) The dependent variable is usually called Y.
(c) It is useful in developing a forecast of one variable as a function of one or more other
variables.
(d) It is helpful in determining what is cause and what is effect.
(e) none of the above
ANSWER: d

5.51

As one increases the number of periods used in the calculation of a moving average,
(a)
(b)
(c)
(d)
(e)

greater emphasis is placed on more recent data.

less emphasis is placed on more recent data.
the emphasis placed on more recent data remains the same.
it requires a computer to automate the calculations.
one is usually looking for a long-term prediction.

ANSWER: b
5.52

The diagram below illustrates data with a

(a) negative correlation coefficient.

(b) zero correlation coefficient.
(c) positive correlation coefficient.

125

Forecasting l CHAPTER 5

(d) correlation coefficient equal to +1.

(e) none of the above
ANSWER: c
5.53

A correlation coefficient of -1 implies that

(a)
(b)
(c)
(d)
(e)

both variables increase at exactly the same rate.

one variable increases at the same rate that the other variable decreases.
the two variables have no correlation.
both variables decrease at the same rate.
both variables increase at the same rate.

ANSWER: b
5.54

If computing a causal linear regression model, Y = a + bX, and the resultant r 2 is very near zero,
then one should conclude that
(a)
(b)
(c)
(d)
(e)

Y = a + bX is a good forecasting method.

a time-series model would be preferable.
a multiple linear regression model would be preferable.
an exponential smoothing model would be preferable.
one's choice of independent variable was inappropriate.

ANSWER: e
5.55

Enrollment in a particular class for the last four semesters had been 120, 126, 110, and 135. The
best forecast of enrollment next semester, based on a 3-semester moving average, would be
(a)
(b)
(c)
(d)
(e)

126.
135.
120.
123.
125.

ANSWER: d

126

Forecasting l CHAPTER 5

5.56

The correlation coefficient resulting from a particular regression analysis was 0.25. What was the
slope of the regression line?
(a)
(b)
(c)
(d)
(e)

0.5
-0.5
0.0625
There is insufficient information to answer the question.
none of the above

ANSWER: d
5.57

A tracking signal was calculated for a particular set of demand forecasts. This tracking signal was
negative. This would indicate that
(a)
(b)
(c)
(d)
(e)

the trend portion of the model was inappropriate.

the nontrend portion of the model was inappropriate.
the EMSE is negative.
the MAD is negative.
none of the above

ANSWER: a
5.58

Given that the MAD for the following forecast is 2.5, what is the actual value in period 2?
Period
1
2
3
4
(a)
(b)
(c)
(d)
(e)

Forecast
100
110
120
130

Actual
95
123
130

120
98
108
115
none

ANSWER: c
5.59

Given that the MSE for the following forecast is 9.5, what is the forecast value in period 3?
Period
1
2
3
4

Forecast
100
110
130

Actual
95
108
123
130

127

Forecasting l CHAPTER 5

(a)
(b)
(c)
(d)
(e)

108
118
128
115
none of the above

ANSWER: e
5.60

Assume that you have tried three different forecasting models. For the first, the MAD = 2.5, for
the second, the MSE = 10.5, and for the third, the MAPE = 2.7. We can then say:
(a)
(b)
(c)
(d)
(e)

the third method is the best.

the second method is the best.
methods one and three are preferable to method two.
method two is least preferred.
none of the above

ANSWER: e
*5.61

Which of the following is not a problem with moving average models?

(a)
(b)
(c)
(d)
(e)

larger number of periods may smooth out real changes

they take a considerable period of time to construct
they dont pick up trends in time to react to the trends
they require that lots of past data be kept
none of the above

ANSWER: b
*5.62

In picking the smoothing constant for an exponential smoothing model, we should look for a value
which
(a)
(b)
(c)
(d)
(e)

produces a nice looking curve.

produces the values you would like to see.
produces values which compare well with actual values based on a standard measure of error.
cause the least computational effort.
none of the above

ANSWER: c

128

Forecasting l CHAPTER 5

*5.63

Which of the following model types would likely be the best for predicting the number of
automobiles sold next year?
(a)
(b)
(c)
(d)
(e)

Delphi
Sales force composite
Regression
Multiple regression
none of the above

ANSWER: d
*5.64

For which of the following forecasts would you expect it to be most appropriate to use a multiple
regression model?
(a)
(b)
(c)
(d)
(e)

The likelihood of a student applicant being accepted at college.

The number of ice cream cones sold by a single store tomorrow.
The number of apples on a single apple tree at harvest time.
The average GPA in a class of seniors graduating from college.
None of the above

ANSWER: a
*5.65

When San Diego Hospital forecast the number of patient days for each upcoming month, they used
a simple regression model. Had they needed to forecast the number of available beds by day for
the upcoming months,
(a)
(b)
(c)
(d)
(e)

a simple regression model would have been more than adequate.

a moving average model would have been more appropriate.
a multiple regression model should have been used.
an exponential smoothing model would have been best.
none of the above

ANSWER: c

PROBLEMS
5.66

For the data below, develop a 3-month moving average forecast.

Month
January
February
March
April
May
June

Automobile
Battery Sales
20
21
15
14
13
16

Month
July
August
September
October
November
December

129

Automobile
Battery Sales
17
18
20
20
21
23

Forecasting l CHAPTER 5

ANSWER:

Month
January
February
March
April
May
June
July
August
September
October
November
December
January
5.67

Automobile
Battery Sales
20
21
15
14
13
16
17
18
20
20
21
23
-

3-Month
Moving Avg.
18.67
18.67
14
14.33
15.33
17
18.33
19.33
20.33
21.33

Use exponential smoothing with = 0.2 to forecast the battery sales. Assume the forecast for
January was 22 batteries.
Month
January
February
March
April

Automobile
Battery Sales
20
21
15
14

ANSWER: Forecasts for January to April are 22, 21.6, 21.48, 20.184
5.68

Use the sales data given below to determine:

(a) the least squares trend line
(b) the predicted value for 1982 sales
Year
1975
1976
1977
1978

Sales (units)
100
110
122
130

Year
1979
1980
1981
1982

Sales (units)
139
152
164
?

ANSWER:
(a) the least squares trend line Y = -20567.4 + 10.46X
(b) the predicted value for 1982 sales = 172.828
5.69

City government has collected the following data on annual sales tax collections and new car
registrations:

130

Forecasting l CHAPTER 5

Annual Sales
Tax Collections
($ millions)
1.0
1.4
1.9
2.0

New Car
Registrations
(thousands)
10
12
15
16

Annual Sales
Tax Collections
(millions)
1.8
2.1
2.3

New Car
Registrations
(thousands)
14
17
20

(a) Determine the least squares regression equation.

(b) Using the results of part (a), find the estimated sales tax collections if new car registrations
total 22.
(c) Calculate the coefficient of correlation.
(d) Calculate the coefficient of determination.
ANSWER:
(a) least squares equation Y = 0.10 + 0.126X
(b) Y = 0.10 + 0.126(22) = 2.67
(c) coefficient of correlation = 0.97461
(d) coefficient of determination = 0.94987
5.70

Let us hypothesize (imagine) that the number of automobile accidents in a certain region are related
to the regional number of registered automobiles in thousands (b1), alcoholic beverage sales in
$10,000 (b2), and decrease in the price of gasoline in cents (b3). Furthermore, imagine that the
regression formula has been calculated as:
Y = a + b1 X1 + b2 X2 + b3 X3
where Y = the number of automobile accidents, a = 7.5, b1 = 3.5, b2 = 4.5, and b3 = 2.5
Calculate the expected number of automobile accidents under the following conditions:
(a)
(b)
(c)

X1
2
3
4

X2
3
5
7

X3
0
1
2

ANSWER:
(a) 28
(b) 43
(c) 58
5.71

Calculate (a) MAD, (b) MSE, and (c) MAPE for the following forecast versus actual sales figures.
Forecast
100
110
120

Actual
95
108
123

131

Forecasting l CHAPTER 5

130

130

ANSWER:
(a) MAD = 10/4 = 2.5
(b) MSE = 38/4 = 9.5
(c) MAPE = (0.0956/4)100 = 2.39%
5.72

Demand for a particular type of battery fluctuates from one week to the next. A study of the last 6
weeks provides the following demands (in dozens): 4, 5, 3, 6, 7, 8 (last week).
(a) Forecast demand for the next week using a 2-week moving average.
(b) Forecast demand for the next week using a 3-week moving average.
ANSWER:
(a) (7+8)/2 = 7.5
(b) (6+7+8)/3 = 7

5.73

Daily high temperatures in the city of Houston for the last week have been as follows:
93, 94, 93, 95, 96, 88, 90 (yesterday).
(a) Forecast the high temperature today using a 3-day moving average.
(b) Forecast the high temperature today using a 2-day moving average.
(c) Calculate the mean absolute deviation based on a 2-day moving average.
ANSWER:
(a) (96+88+90)/3 = 91.3
(b) (88+90)/2 = 90
(c) MAD = 13.5/5 = 2.7

5.74

Average starting salaries for students using a placement service at a university have been steadily
increasing. A study of the last four graduating classes indicate the following average salaries:
$20,000, $22,000, $23,000, and $25,000 (last graduating class).
Predict the starting salary for the next graduating class using an exponential smoothing model with
= 0.2. Assume that the initial forecast was $20,000 (so that the forecast and the actual were the
same).
ANSWER: Forecast for next period = 21736

5.75

A firm conducted a careful analysis of the cost of operating an automobile. The following model
was developed:
Y = 4000 + 0.20X, where Y = annual cost, X = miles driven
(a) If a car is driven 15,000 miles this year, what is the forecasted cost of operating this
automobile?

132

Forecasting l CHAPTER 5

(b) If a car is driven 25,000 miles this year, what is the forecasted cost of operating this
automobile?
(c) Suppose that one car was driven 15,000 miles and the actual cost of operating was $6,000,
while a second car was driven 25,000 miles and the actual operating cost was $10,000.
Calculate the mean absolute deviation for this.
ANSWER:
(a) 4000 + 0.20(15000) = 7000
(b) 4000 + 0.20(25000) = 9000
(c) MAD = (1000 + 1000)/2 = 1000
5.76

The following multiple regression model was developed to predict job performance as measured by
a company job performance evaluation index based on a pre-employment test score and college
grade point average (GPA).
Y = 35 + 20X1 + 50X2, where Y = job performance evaluation index
X 1 = pre-employment test score
X 2 = college GPA
(a) Forecast the job performance index for an applicant who had a 3.0 GPA and scored 80 on the
pre-employment score.
(b) Forecast the job performance index for an applicant who had a 2.5 GPA and scored 70 on the
pre-employment score.
ANSWER:
(a) 35 + 20(80) + 50(3.0) = 1785
(b) 35 + 20(70) + 50(2.5) = 1560

133

Forecasting l CHAPTER 5

5.77

Given the following data, if MAD = 1.25, determine what the actual demand must have been in
period 2 (A2).
Time Period
1
2
3
4

Actual (A)
2
A2 = ?
6
4

Forecast (F)
3
4
5
6

|FA|
1
1
2

ANSWER:
A2 = 3 or A2 = 5
*5.78

For the data below, develop a 3-month moving average forecast.

Month
January
February
March
April
May
June

Automobile
Tire Sales
80
84
60
56
52
64

Month
July
August
September
October
November
December

Automobile
Tire Sales
68
100
80
80
84
92

ANSWER:

Month
January
February
March
April
May
June
July
August
September
October
November
December
January
*5.79

Automobile
Tire Sales
80
84
60
56
52
64
68
100
80
80
84
92
-

3-Month Tire
Avg.
74.7
66.7
56.0
57.3
61.3
77.3
82.7
86.7
81.3
74.7

Use exponential smoothing with = 0.3 to forecast the battery sales. Assume the forecast for
January was 22 batteries.

134

Forecasting l CHAPTER 5

Month
January
February
March
April

Automobile
Tire Sales
20
21
15
14

ANSWER:
Automobile
Month
Tire Sales Forecasts
January
20
22.00
February
21
19.40
March
15
21.48
April
14
13.06
*5.80

Use the sales data given below to determine:

(a) the least squares trend line
(b) the predicted value for 1982 sales
Year
1995
1996
1997
1998

Sales (units)
130
140
152
160

Year
1999
2000
2001
2002

Sales (units)
169
182
194
?

ANSWER:
(a) the least squares trend line Y = -20746.6
(b) the predicted value for 2002 sales = 202.8571

135

Forecasting l CHAPTER 5

*5.81

For the data below

Year
1980
1981
1982
1983
1984
1985
1986

Automobile
Sales
116
105
29
59
108
94
27

Year
1987
1988
1989
1990
1991
1992
1993

Automobile
Sales
119
34
34
48
53
65
111

(a) Develop a 6-year moving average forecast.

(b) Find the MAD.
ANSWERS:
Year
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993

Number of
Automobiles Forecast
116
X
105
X
29
X
59
X
108
X
94
X
27
85.2
119
70.3
34
72.7
34
73.5
48
69.3
53
59.3
65
52.5
111
58.8

Error

|Error|

-58.2
48.7
-38.7
-39.5
-21.3
-6.3
12.5
52.2

58.2
48.7
38.7
39.5
21.3
6.3
12.5
52.2

MAD = 34.67

SHORT ANSWER/ESSAY
5.82

In general terms, describe what time-series forecasting models are.

ANSWER: forecasting models that make use of historical data

5.83

In general terms, describe what causal forecasting models are.

ANSWER: forecasting models that incorporate variables or factors that might influence the
quantity being forecasted

5.84

In general terms, describe what qualitative forecasting models are.

ANSWER: forecasting models that attempt to incorporate judgmental or subjective factors into the
model

136

Forecasting l CHAPTER 5

5.85

Briefly describe the Delphi forecasting method.

ANSWER: an iterative group process that allows for individuals, at various locations, to make a
qualitative forecast

5.86

Briefly describe the executive opinion forecasting method.

ANSWER: a forecasting model that uses the opinions of a small group of high-level managers and
makes a group estimate of demand

5.88

Briefly describe the sales force composite forecasting method.

ANSWER: a forecasting method in which each person is responsible for a regions estimate of
sales and the estimates are combined after review to make a district/country/continent overall
forecast

5.89

Briefly describe the consumer market survey forecasting method.

ANSWER: a forecasting method that solicits input from customers or potential customers
regarding their future purchasing plans

5.90

List the possible components of time-series data.

ANSWER: trend, seasonality, cycles, random variations, or noise

5.91

In general terms, describe an independent variable.

ANSWER: a set of variables that potentially may be useful in forecasting a dependent variable

5.92

In general terms, describe a dependent variable.

ANSWER: the variable being forecasted

5.93

The accuracy parameter of regression estimates is called the ________________.

ANSWER: standard error of the estimate

5.94

List four measures of historical forecasting errors.

ANSWER: MAD, MSE, MAPE, Bias

*5.95

In what way might it be said that all forecasting models are subjective?
ANSWER: The forecaster always make a choice as to which variables to include, what data is
acceptable, which measurements appropriate, etc.

*5.96

Explain, briefly, why most forecasting error measures use either the absolute or the square of the
error.
ANSWER: A deviation is equally important whether it is above or below the actual. This also
prevents negative errors from canceling positive errors which would understate the true size of the
errors

137

Forecasting l CHAPTER 5

*5.97

Explain, briefly, why the larger number of periods included in a moving average forecast, the less
well the forecast identifies rapid changes in the variable of interest.
ANSWER: The larger the number of periods included in the moving average forecast, the less the
average is changed by the addition or deletion of a single number.

*5.98

Explain, briefly, why, in the exponential smoothing forecasting method, the larger the value of the
smoothing constant, , the better the forecast will be in allowing the user to see rapid changes in
the variable of interest.
ANSWER: The larger the value of , the greater is the weight placed on the most recent
occurrences.

*5.99

Explain, briefly, why the Delphi forecasting approach is probably the most useful of those
discussed when attempting to forecast fifty to one hundred years into the future.
ANSWER: (a) there is little or no useful quantitative data available, (b) the knowledge of several,
perhaps many, disciplines will be required, (c) you wish to be able to weight each of the sources
about equally.

*5.100 The decomposition approach to forecasting (using trend and seasonal components) may be helpful
when attempting to forecast a time-series. Could an analogous approach be used in multiple
regression analysis? Explain, briefly.
ANSWER: An analogous approach would be possible using time as one independent variable and
using a set of dummy variables to represent the seasons.
*5.101 What are some of the basic assumptions we make when using simple linear or multiple regression?
ANSWER: (1) independent errors, (2) random error, of mean zero, and more or less normally
distributed in nature.
*5.102 What is one advantage of using causal models over time-series or qualitative models?
ANSWER: Use of the causal model requires that the forecaster gain an understanding of the
causes, not merely the frequency of variations, i.e., gains a greater understanding of the problem,
than the other methods.

138