Chapter 5

1. Explain the concept of time series decomposition.

Each component is identified separately. The projections of each of the
components can then be used in combination to make forecasts of
future values of the time series. Decomposition methods are used to
forecasts for both short and long terms. They are also used to display of
a simple way to understand the underlying growth or decline of a series, as well as for
adjust the series by removing one or more of the components.
2. Explain when a multiplicative decomposition is more suitable than a
additive decomposition.
The multiplicative model works better when the variability of the series increases with the
level, that is, the values of the series scatter as the trend increases
3. What are some basic factors that influence the trend cycle of the
majority of the variables?
Population growth, inflation, product sales in their growth stage
the life cycle
4. What kind of trend model should be used in each of the
next cases?
a) The variable increases at a constant rate.
It is possible to fit an exponential trend.
b) The variable increases at a constant rate until it reaches saturation and levels off.
Non-linear trend
c) The variable increases by a constant amount.
Non-linear trend curves
5. What are some basic factors that influence the seasonal component of
most of the variables?
School periods, vacation periods, seasonal products, seasons of the year
 6. The sales and revenue growth estimates of Valué Line for companies
individuals are derived from the correlations of sales, income, and dividends of the
appropriate components of National Income Accounts, such as investments in
capital goods. Jason Black, an analyst from Value Line, is reviewing the
trend of the variable of investments in capital goods from 1977 to 1993. The
data is presented in table P-6.
As a first step, we need to open the Minitab program and this is the main screen of the
program

After opening, we proceed to enter the data and then we perform an analysis.
descriptive

a) Descriptive analysis of the box plot and compliance with assumptions and the series is
stationary or not stationary
Statistics
Error
standard
of the
Variable N N* Media media Minimum Standard Deviation Q1 Median Q3 Maximum
miles 17 0 433.5 29.9 123.4 214.0 340.0 437.0 561.0 623.0
1. 4. 5. 6. 7. 8. 9. 10 11.
In the descriptive analysis

The data has an average of 433.5 and a maximum of 623.

at least 214 and the data is 17 in total.
 Box diagram

Interpretation: there are no outliers in the series.

Analysis of the components of the series

It is not autocorrelated because it is not.

As we can see in the graph we can
to say that it has an upward trend surpasses no backlog therefore
we can say that it is stationary
a) Graph the data and determine the appropriate trend model for the years of
1977 to 1993.

As we can see from the graph, the appropriate model for this data is the linear trend.
with Yt = 218.5 + 23.89*t

b) If the appropriate model is linear, calculate the linear trend model for the years
from 1997 to 1993.

We applied the linear trend model since it was the best fit for our data.
with a MAPE of 4.502
c) What is the average annual increase in investments in capital goods since
1977?

The average annual increase is above the average by 22%

d) Estimate the value of the trend for investments in capital goods in 1994.

the forecast for the year 1994 for investments is 648,581

e) Compare your trend estimate with that of Value Line.

when the comparison is made, we can see the actual data with the forecast there is a
big difference this means that it was not a better model for prediction

What factor(s) influence the trend of investment in capital goods?

The factor that influences the trend of investment in capital goods is inflation.

The assumptions

As can be seen in the graph, there is normality; there is a random variable because they are.
dispersed and it is not very clear that it has homoscedasticity because they do not follow a pattern
 7.A large American company is considering making cuts to its
advertising on TV and instead, deliver business videos to their clients. This
The action is being reviewed after the company's president read.
recently an article in the newspaper referring to videos to attract
clients as 'the weapon to achieve sales' nowadays. Something that the president
would like to research before undertaking this action is the history of advertising
on TV in the United States, especially in relation to the trend cycle. The
Table P-7 indicates the total spending in dollars on television advertising in the United States
(in millions of dollars).

YEAR Y
1980 11,424
1981 12,811
1982 14,566
1983 16,542
1984 19,670
1985 20,770
1986 22,585
1987 23,904
1988 25,686
1989 26,891
1990 29,073
1991 28,189
1992 30,450
1993 31,698
1994 35,435
1995 37,828
1996 42,484
1997 44,580

Statistics
Variable Median Q3
Y 18 0 26366 9704 94171989 36.81 474586 11424 18888 26289 32632
N for
Variable
Y 44580 33156 * 0.27
0 -0.60
Box diagram

Interpretation:

The data has a mean of 26.366 and a

median of 26, 289 does not have a mode and its
desviación de 97,04 y un máximo es de 44,558
and at least 11,424 and the data is 18 in
total. As can be seen in the graph of
box diagram no outlier data exists

It is not autocorrelated because it is not.

does not surpass any backlog therefore
we can say that it is stationary

a) Graph the time series of television advertising expenses in the States

United.

In the following chart, we observe that the

the time series is increasing

The advertising expenditure data shows a growing trend over time.

b) Fit a linear trend to the advertising data and plot the fitted line.
about the time series graph.

Adjusted trend equation

Yt = 9310 + 1795.4×t
 The data that fits best is a linear trend with an equation of
y=9310+1795.4×te and with a MAPE error of 3

c) Forecast the television advertising expenses in dollars for 1998.

Forecasts
Forecast
19 40.0818 29992.7 50170.8

The advertising spending forecast in the United States for the year 1998 is
40,818 that means they are going to decrease for that year compared to the previous year

It can be observed in the graph of the assumptions if it meets the assumptions of the
normality and also with homoscedasticity as well as independence since the
data is scattered
 d) According to the results of the section, do you believe there could be a component
cyclical in spending on television advertising in dollars? Explain.

Yes, since this component reflects recurring behaviors, although they do not have
Why be exactly newspapers, with a period longer than a year. They show,
Usually, how do periods of economic prosperity succeed those of crisis?
or at least, deceleration.

8. Suppose that the following specific seasonal indices for March are given
as percentages and were obtained by the average proportion method
mobile

102.2 105.9 114.3 122.4 109.8 98.9

What is the seasonal index for March using the median?
102.2+105.9+114.3
+122.4+109.8+98.9=108.917

The seasonal index using the mean was: 108.917

9. The expected value of the trend for October is $850. Assuming an index
seasonal for October of 1.12 (112%) and the multiplicative model given by the
Equation 5.2, what would be the forecast for October?

The expected value of the trend for October applying the multiplicative model
with the following equation
Y=1.12+5.2*t Y=1.12+5.2*850 = 957.2
The expected value of the trend for October would be: 975.2
10. The following specific percentages for the seasonal indexes correspond to
to the month of December:
75.4 86.8 96.9 72.6 80.0 85.4

Assume a multiplicative decomposition model. If the expected trend for

December is $900 and the seasonal adjustment of the median is used, what is the
forecast for December?
 75.4+86.8+96.9+72.6+80.0+85.4
=82.85
6

xt = Tt + St + at
xt = 900 + 82.85 + 1 = 983.85

The forecast for December is: 983.85

A large resort near Portland, Maine, has kept a record of its
monthly sales over several years, but has never analyzed this data. The
the vacation center calculates the seasonal indices for its monthly sales.
Which of the following statements about the index are correct?
a) The sum of the 12 monthly index numbers, expressed as percentages,
It should be 1,200.

I=12/1200=0.01 is Incorrect
An index of 85 for May indicates that sales are 15% lower than sales.
average monthly.
Y=0.15+0.01*85=1
Correct
In preparing a report for June Bancock, manager of Kula Department Store,
you include the sales figures from the last year (in thousands of dollars) presented in the
Table P-12. Upon seeing them, Mr. Bancock says: "This report confirms what I have been
saying: the business is getting better every time. Is this statement correct? Why?

Index
seasonal
Me Sales adjusted
January 125 51
February 113 50
March 189 87
April 201 93
May 206 95
June 241 99
July 230 96
August 245 89
September 271 103
October 291 120
November 320 131
December 419 189
As we can see in the graph, we can say that the business is improving more and more and
the statement is correct because the business is growing
12. The quarterly sales figures (measured in millions of dollars) of Goodyear
These are presented in table P-13. Is there a significant seasonal effect?
at these sales levels? Analyze this time series to obtain the four
seasonal indices and determine the magnitude of the seasonal component in the
Goodyear sales.
 In the graph, we can see that it tends to grow positively.

Interpretation

The data does not have typical or missing data like

its average is 2.808, with a minimum of 2.063 and with
a maximum of 3,351 and no data found
atypical

If it is autocorrelated because if
It overcomes delays, therefore we can say
what is stationary

ANALYSIS OF THE SERIES

As can be seen in the graph, there is a growing trend and there is also
seasonality and the random factor
a) To forecast, would you use the trend component, the component of the
seasonality or both?

To forecast these data, it is recommended to use both components as

both fit

b) Prepare a forecast for the third and fourth quarters of 1996.

Forecasts
Forecast Period
47 3.250
48 3.452
 The forecast for the third period is 3,250 and for the fourth
the first quarter of 1996 is 3.452

c) Compare your forecasts with those of Value Line.

Forecasts
Forecast Period
47 3.250
48 3.452

The forecasted values have a difference of decimals compared to the actual value since
the model that was chosen was not the most suitable
13. The monthly sales of Cavanaugh Company, represented in figure 5.1
(below), they are specified in table P-14.
The series has a trend and is growing positively

Statistics
Count
Variable total Mean Minimum Q1 Median Q3
Sales_2 77 298.4 198.4 39374.7 66.50 22977.0 36.0 146.5 257.0 398.0
N for
Variable mode
Ventas_2 895.0 859.0 169, 210, 223, 2 1.04 0.61
272

Interpretation

If the data has typical data, its mean is

298.4 at a minimum is 36 and with a maximum of 895
its average of 257 and its mode of 2 its skewness
of 1.04 and a kurtosis of 0.61
 Interpretation

If it is autocorrelated, that is, if

it is stationary since it exceeds 3
lags of the confidence line

We observe that it has no trend, and it has seasonality and a random factor.

a) Perform a multiplicative decomposition of the time series of sales

Cavanaugh Company, assuming the components of trend, seasonal and
irregular.

MULTIPLICATIVE DECOMPOSITION

Method
Type of model Multiplicative model
Data Sales_2
 Length 77
Number of values 0
missing

Adjusted trend equation

Yt = 72.6 +
6.013×t

Seasonal indices
Period Index
1 1.27806
2 0.90740
3 0.61574
4 0.48172
5 0.42562
6 0.46744
7 0.65326
8 0.86343
9 1.36478
10 1.78981
11 1.86506
12 1.28769

Measures of accuracy
MAPE 14.21
MAD 32.63
MSD 1913.95

Interpretation:
We observe in the graph that if the multiplicative model is adjusted, we also say that,
if there is normality, homoscedasticity, and independence.

b) To forecast, would you use the trend component, the seasonal component?
or both?

For the forecast, I would use the seasonal component as it fits better.
data. Many economic series show regular fluctuations in the same month
of each year, and with some guidelines that are presented, without repeating exactly, all
the years.
c) Obtain forecasts for the rest of 2006.
For the rest of 2006, the forecasts are:

Forecasts
Forecast Period
78 253.17
79 357.74
80 478.03
81 763.80
82 1012.44
83 1066.21
84 743.89

14. Build a table similar to table P-14 with the natural logarithms of the
monthly sales. For example, the value for January 2000 is ln(154) = 5.037.
 Sales Sales Sales Sales Sales Sales Sales
Year 2000
January 5.037 5.298 5.407 5.846 6.250 6.418 6.443
February 4.564 4.771 4.644 5.565 6.001 5.971 5.730
March 4.290 4.500 4.673 5.412 5.704 5.609 5.781
April 3.892 4.369 4.443 4.949 5.347 5.775 5.513
May 3.584 4.357 4.317 4.997 5.278 5.242 5.606
June 4.078 4.511 4.595 4.977 5.226 5.549
July 4.554 5.118 4.905 5.407 5.509 5.781
August 5.130 5.130 5.352 5.606 5.838 6.001
September 5,347 5.666 5.814 6.098 6.140 6.518
October 5.628 5.849 6.131 6.328 6.522 6.755
November 5.697 5.927 6.190 6.417 6.567 6.797
December 5.501 5.313 5.787 6.146 6.413 6.498

Statistics
Count Deviance Est Minimum
Variable total Media. Variance o Q1 Median
Sales 77 5.4662 0.7212 0.5202 13.19 420.899 3.5835 4.987 5.5491
(ln) 8 0
N for
Variable Q3 Maximum Range Mode mode asymmetry kurtosis
Sales 5.9863 6.7968 3.2133 5.12990 5.34711 -0.36 -0.39
(ln) 5.40717
5.60580
 Interpretation

The data does not have typical data, its average is

5.466 as a minimum is 3.58 and with a maximum of
6.7968 on average of 5.5491 and its mode of 2
asimetría de -0.36 y una curtosis de -0.39

Interpretation

If it is autocorrelated, that is to say, if it is

stationary since it exceeds 3 delays from the line
reliable

a) Make an additive decomposition of ln(sales), considering the model.

Y=T+S+I
Method
Type of model Additive model
Data Sales (ln)
Length 77
Number of values 0
missing

Seasonal indices
Period Index
1 0.334619
2 0.018136
3 0.402487
4 0.636992
5 0.714009
6 0.570579
7 0.273000
 8 0.001204
9 0.469961
10 0.722911
11 0.746716
12 0.342200

Measures of accuracy
MAPE 8.52519
MAD 0.44378
MSD 0.25754

c) To forecast, would you use the trend component, the component

seasonal or both?

To forecast, it is recommended to use seasonal decompositions since

As we observe in the graph, the green line represents this decomposition of the
trend and we observe that it is constant.
c) Develop forecasts for ln(sales) for the remaining months of 2006.
The forecasts for the remaining months are:

Forecasts
Forecast Period
78 4.91432
79 5.21189
80 5.48369
81 5.95485
82 6.20780
 83 6.23161
84 5.82709

d) Take the antilogarithms of the forecasts calculated in the item to obtain

forecasts of actual sales for the rest of 2006.
For the rest of 2006, the forecasts are:
Forecasts
Forecast Period
78 253.17
79 357.74
80 478.03
81 763.80
82 1012.44
83 1066.21
84 743.89

Forecasts
Forecast Period
78 4.91432
79 5.21189
80 5.48369
81 5.95485
82 6.20780
83 6.23161
84 5.82709
We observe that forecasts without algorithms are the best as they provide us
data that matches the real data.
d) Compare the forecasts from the subsection with those from the subsection of the problem.

We can say that the actual forecasts are more present since when it is applied
Logarithms result in smaller outcomes.
Which set of forecasts do you prefer? Why?
quarterly sales (in millions of dollars) of Disney Company from the first
from the third quarter of 1980 to the third quarter of 1995.
 These data show a seasonal pattern. The pattern repeats every 12 months.

Statistics
Medi Count
Total variable a Standard Deviation
ventas_3 63 1134 871 758644 71464 204 363 775 1739
N for
Variable
sales_3 3302 3098 * 0.83 -0.41
0
 Interpretation

The data does not have typical values; its mean is

1134 at least is 204 and with a maximum of
3302 its average of 775 and its mode of 0
asimetría de 0.83 y una curtosis de -0.41

Interpretation

If it is autocorrelated that is to say if it is

stationary since it exceeds 4 lags of the line
reliable

a) Make a multiplicative decomposition of the time series that integrates the

Disney's quarterly sales.

Method
Type of model Multiplicative model
Data sales_3
Length 63
Number of values 0
missing
Adjusted trend equation
Yt = -302.9 +
44.89×t
Measures of accuracy
MAPE 37.0
MAD 224.2
MSD 79278.5
b) Is there a significant trend? Discuss the nature of the component.
seasonal.

If there is a growing trend in the seasonal component since many economic series
they present regular oscillations in the same month each year, and with patterns that
they present, without repeating exactly, every year.
Would you use both components, the trend and the seasonal, for forecasting?
I would only use the one with the trend since we observe that there is a trend and not seasonality.

d) Forecast the sales for the fourth quarter of 1995 and the four quarters of
1996
The forecasts for the fourth quarter of 1995 and the other four quarters of 1996 are:

Forecasts
Forecast Period
64 2506.18
 65 2501.53
66 2719.25
67 2830.03
68 2681.28

16. The monthly gasoline demand (in thousands of barrels/day) from Yukong Oil
South Korean company for the period from January 1986 to September
from 1996 is indicated in table P-17.

Time series chart of SALES

3500

3000

2500

2000

1500

1000

500

1 6 12 18 24 30 36 42 48 54 sixty
Index

In this graph we observe multiplicative seasonal changes, the magnitude of the

seasonal changes increase over time as data values
are increasing.
Statistics
Count
Total variable Mean Minimum Q1 Median
Sales4 129 84.21 53.58 2870.67 10863.67 0.04 34.65, 74.10
N for
Variable Q3 Maximum
ventas4 124.75 216.10 216.06 68.1 3 0.47 -0.88

Interpretation

The data does not have typical data; its average is

84.21 at a minimum is 0.04 and with a maximum of
21.10 with an average of 74.10 and a mode of 3.
asimetría de 0.47 y una curtosis de -0.88

Autocorrelation function for SALES

(with significance limits of 5% for the autocorrelations)
1.0

Interpretation
0.8

0.6

0.4

0.2 If it is autocorrelated that is, if it is

0.0

-0.2 stationary since it exceeds 7 laggings of the line

-0.4

-0.6
trustworthy
-0.8

-1.0
1 5 10 15 20 25 30 35 40 45 50 55 60
Phase shift

a) Graph the time series of gasoline demand. Do you think it would be

an additive or a multiplicative decomposition suitable for this series of
time? Explain.
It can be observed that the multiplicative decomposition is better since it fits better.
the data to the series that we have

b) Conduct an analysis of the decomposition of gasoline demand.

Method
Type of model Multiplicative model
Data sales4
Length 129
Number of values 0
missing

Adjusted trend equation

Yt = -6.37 +
1.3914×t

Seasonal indexes
Period Index
1 0.94408
2 0.94698
3 0.95790
4 0.99406
5 1.00071
6 1.01123
7 1.02689
8 1.07703
9 1.04103
 10 0.98261
11 0.99809
12 1.01939

Measures of accuracy
MAPE 1807.64
MAD 8.43
MSD 122.48

c) Interpret the seasonal indices.

Seasonal indices
Period Index
1 0.94408
2 0.94698
3 0.95790
4 0.99406
5 1.00071
6 1.01123
7 1.02689
8 1.07703
9 1.04103
10 0.98261
11 0.99809
12 1.01939
We observe that the multiplicative decomposition is the best because it fits better.
the series.
d) Forecast the demand for gasoline for the last three months of 1996.
 The last three months of 1996 are:

Forecasts
Forecast Period
130 171.485
131 175.575
132 180.740

17. Table P-18 contains data representing monthly sales (in thousands of
millions of dollars) from all retail stores in the United States. Based on
In the 1994 data, perform a decomposition analysis of this series. Make
comments about the three components of the series. Forecast the sales to
retail for 1995 and compare its results with the actual values that appear
on the table.

Gasoline time series graph

200

150

100

50

0
1 13 26 39 52 65 78 91 104 117
Index

The series is seasonal as it is growing positively.

Statistics
Count
Variable total Mean Minimum Q1 Median
P18 84 157.70 22.54 508.14 14.29 13247.10 113.60 142.50 154.35
N for
Variable Q3 Maximum Range Mode fashion Kurtosis Asymmetry
P18 174.55 233.30 119.70 130.9, 145, 154.6, 2 0.65 0.78
164.6

Interpretation

If the data has typical data, its average is of

157.70 at least is 113.60 and with a maximum of
233.30 its average of 154.35 and its mode of 2 its
asimetría de 0.65 y una curtosis de 0.78

Interpretation

If it is autocorrelated, that is, if it is

stationary since it exceeds 5 delays of the line
trustworthy
We observe that it has no trend, but it is seasonal and has a random variable.
Multiplicative decomposition with trend and seasonality

Method
Type of model Multiplicative model
Data P18
Length 84
Number of values 0
missing

Adjusted trend equation

Yt = 128.814 +
0.6767×t

Seasonal indices
Period Index
1 0.88020
2 0.85927
3 0.99140
4 0.98614

5 1.03112
6 1.02073
7 1.00741
8 1.03463
9 0.97311
10 0.99059
11 1.01547
12 1.20993

Measures of accuracy
MAPE 1.9842
MAD 3.1934
MSD 16.5349

Forecasts
Forecast Period
85 164.010
86 160.690
87 186.071
88 185.751
89 194.920
90 193.648
91 191.802
92 197.684
93 186.590
94 190.611
95 196.085
96 234.453
The multiplicative decomposition fits the data better and therefore we say that it is the
better if there is normality, homoscedasticity, and independence.

1995 forecast of retail sales are:

Forecasts
Forecast Period
85 164.010
86 160.690
87 186.071
88 185.751
89 194.920
90 193.648
91 191.802
92 197.684
93 186.590
94 190.611
95 196.085
96 234.453

18. The seasonally adjusted indices that appear in table P-19 reflect the
changing volume of business at the Mount Spokane Resort Hotel, which offers
services to families in the summer and to enthusiastic skiers during the months of
winter. No drastic cyclical variations are expected during 2007.

Statistics
Count
Variable total Mean
index 12 100.0 39.4 1550.9 39.38 1200.0 33.0 65.5 98.5 134.0
 N for
Variable
index 153.0 120.0 * 0 -0.29 -0.90

Interpretation
The data does not have typical values; its average is
At least 100 is 33 and a maximum of 153 its
promedio de 98.5 y su moda de 0 su asimetría de -
0.29 and a kurtosis of -0.90

Interpretation
It is not autocorrelated, that is to say, it is not
stationary as it does not exceed any backlog of
the line of trust

a) If 600 tourists stayed at the hotel in January 2007, what would be a

reasonable estimate for February?

We adjusted the series to a multiplicative model and it gives us the following equation. For
being able to forecast the tourists who will stay in February 2017 will be
Y = 113.0 - 1.77 * 600 = 949
 b) The equation of the monthly trend is = + where t = 0 represents the
January 15, 2001. What is the forecast for the month of 2007?
T =140 + 5t ( )
T =145 7= 175
We observe that the forecast for the month of September is 175
c) What is the average number of new tourists per month?
Y = 113.0 - 1.77 * 12 = 91.76 = 92 people

The average number of tourists per month is 92.

19. Analyze the performance of the composite index of the indicators with
forward-looking behavior as a barometer of business activity in
recent years.
This indicator provides forward signals about the possible evolution of the
economy. In other words, based on a series of variables, the CLI attempts to predict what is going to happen.
what will happen in that economy in the next 6-9 months. If the CLI increases, it means that the
The economy is going to maintain sustained growth in the coming times. If it reaches a
ceiling, it means that the economy will be below its potential. And if it falls, the
the economy could enter a recession.
Currently, both the Spanish index and the general OECD index are on the decline, and there has been
situated below 100 points, the reference value, in a clear sign of
deceleration.

20. What is the current position of the business cycle? Is it expanding or

contracting? When will the next turning point be presented?
In itself thehypothesisthe
business cycle emphasizes the sense of the change of the
economy. For politicians to be reelected, the unemployment rate must be
decreasing and the inflation rate should not be worsening.
21. What is the purpose of deflating a time series measured in dollars?
If we have a statistical series of data on the evaluation of some
magnitude economic (consumption , production
etc.), it is common for the monetary valuation of this data to be done at prices
currents of each period. As prices undergo fluctuations of
from one period to another, the series as represented does not allow for comparisons. The
The solution to this problem is to express the series in terms of constant prices.
a certain period (base year).
 22. The P-25 table contains the number (in thousands) of 16-year-old males in
forward that were employed in the United States for the months of January 1993
to October 2003.

Use Minitab to perform a multiplicative decomposition of these data and generate the
forecasts for the next 12 months.

Statistics
Content
Variable Medi Desv. Est Variance CoefVa Minimum Median
e total a . a r Sum or Q1 a Q3
p25 130 7011 2975 885123 4.24 911550 63344 6742 70738 7269
9 2 3 2 7
Maximum Range Variable N for Asymmetry
e o o Fashion way to Kurtosis
p25 74579 1123 * 0 -0.37 -1.01
5

Interpretation

The data does not have typical data, its average is

70119 at a minimum is 63344 and with a maximum of
74579 its average of 70738 and its mode of 0
asimetría de -0.37 y una curtosis de -1.01
 Interpretation

If it is autocorrelated, that is, if it is

stationary as it exceeds 11 lag of the line
trustworthy

We observe that there is a growing trend; we can also say that there is seasonality.
random variable

Multiplicative Model

Method
Type of model Multiplicative model
Data p25
Length 130
Number of values 0
missing
Adjusted trend equation
Yt = 65355 +
72.67 × t
Seasonal indices
Period Index
1 0.98118
2 0.98484
3 0.98973
4 0.99534
5 1.00191
6 1.01375
7 1.01906
8 1.01372
 9 1.00207
10 1.00439
11 0.99889
12 0.99513
Measures of accuracy
MAPE 1
MAD 597
MSD 551164

We observe that the multiplicative model fits the data; we can also say that
there is normality and homoscedasticity and independence.
Do the forecasts seem reasonable?
The forecasts for the next 12 months are:

Multiplicative model forecasts

Forecast Period
131 74791.4
132 74581.7
133 73607.8
134 73954.0
135 74393.4
136 74887.2
137 75454.0
138 76419.5
139 76894.1
140 76564.4
141 75757.2
142 76005.6
Exponential smoothing forecasts
Forecast
131 74059.5 72954.6 75164.3
 132 74059.5 72954.6 75164.3
133 74059.5 72954.6 75164.3
134 74059.5 72954.6 75164.3
135 74059.5 72954.6 75164.3
136 74059.5 72954.6 75164.3
137 74059.5 72954.6 75164.3
138 74059.5 72954.6 75164.3
139 74059.5 72954.6 75164.3
140 74059.5 72954.6 75164.3
141 74059.5 72954.6 75164.3
142 74059.5 72954.6 75164.3

Yes, since it is not constant, on the other hand, the exponential smoothing model is
constants
Exponential smoothing model

Measures of accuracy
MAPE 1
MAD 451
MSD 292767

Is a multiplicative decomposition appropriate for this case? Explain.

If the best since the forecasts are not consistent on the other hand the forecasts of the other
The additive model is constant and the multiplicative model fits the data better.
23. Refer to problem 25. The multiplicative decomposition in Minitab assumes that
default a linear trend. Graph the data for the males
employees from table P-25 and examine the years from 1993 to 2000 and from 2001 to 2003. Do you
Is an appropriate linear trend? If not, can you suggest a curve?
of trend that could be suitable? Adjust your suggested trend curve and
I stored the residuals. I calculated the autocorrelation function of the residuals. Did the auto
Do residual correlations suggest a seasonal component? Explain.
Multiplicative model Quadratic model

We observe that the multiplicative models fit the data better as we observed.
above we observe that there is normality, homoscedasticity, and independence.

24. The table P-27 indicates the quarterly sales (in millions of dollars) of the stores.
Wal-Mart from 1990 to 2004. Use Minitab to perform a multiplicative decomposition.
from the time series of Wal-Mart sales from the years 1990 to 2003 and generate
forecasts for the four quarters of 2004.
 a) Statistics
Variable CoefVa Minimum Median
e total a . Variance r Sum or Q1 a
p27 56 32070 18431 33969140 1690 27968
8 4 1
Maximum N for
Variable Q3 o Range
p27 4807 75190 68422 0 * 0.54 -0.83
1

INTERPRETATION OF STATISTICAL DESCRIPTION

The data is a total of 56 with a minimum of 67.68 and

with a maximum of 75,190 and an average of 32,070 and
In the box plot, it can be seen that there are no data.
atypical

b) INTERPRETATION OF THE COMPONENTS OF THE SERIES

Autocorrelation Partial autocorrelation

As can be seen that the data That they are not autocorrelated and that
they are autocorrelated and that it means they have seasonality
it means that the data has
 c) ANALYSIS OF THE COMPONENTS OF THE SERIES

It can be clearly seen

that has a trend
increasing and a random factor
and it cannot be observed
clearly that has
estacionalidad

Multiplicative Model
Method

Model type Multiplicative model

Data p27
Length 56
Number of values 0
missing
Adjusted trend equation
Yt = 1165 +
1081.4×t
Seasonal indices
Period Index
1 0.92253
2 0.98493
3 0.95995
4 1.13259
 d) Is a multiplicative decomposition suitable for Wal-Mart's data?
Explain. Is there a strong seasonal component? Are you surprised?

The multiplicative decomposition is not very suitable as the forecasted values do not
they fit well to the original data line also the errors are very high but the one that
well adjusted is the simple exponential model with a MAPE error of 11

e) Compare the quarterly forecasts of 2004 with the actual sales. Do the
Do the results reinforce the selection of a multiplicative decomposition?

It can be observed that the actual data with the forecasted data with the
multiplicative decomposition that the forecasts are above the data
real

25. Refer to problem 27. The multiplicative decomposition in Minitab assumes the
by default a linear trend. Fit and plot a line of
linear trend for Wal-Mart's sales. Is the linear trend suitable for
these data? If not, could you suggest a trend line that results
suitable? Adjust your suggested trend curve and save the residuals. I calculated the
auto correlations of the residuals. Do the residual auto correlations suggest a
seasonal component? Explain.
a) Statistics, series graph and box plot
Time series graph of Sales_1
90

80

70

60

50

40

30

20

10

0
1 6 12 18 24 30 36 42 48 54 60
Index

In this graph, we observe that the magnitude of seasonal changes increases

over time as the data values increase.

INTERPRETATION OF
STATISTICAL DESCRIPTION

The data totals 56 with

a minimum of 6.77 and with a
maximum of 82.82 with an average
of 34.73 and in box diagram it
you can see that there is no data
atypical

b) INTERPRETATION OF THE COMPONENTS OF THE SERIES

Autocorrelation Partial autocorrelation

 Autocorrelation function for Sales_1
(with significance limits of 5% for autocorrelations) Partial autocorrelation function of Sales_1
(with significance limits of 5% for the partial autocorrelations)
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0
-0.2
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
-0.8
-1.0
-1.0

1 5 10 15 20 25 30 35 40 45 50 55 1 5 10 15 20 25 30 35 40 45 50 55
Phase shift Phase shift

As can be seen, the data is autocorrelated and that means that the
data have a trend

Is there a strong seasonal component?

It can be clearly observed that it has a

growing trend and a random factor and do not know
you can clearly see that it has
seasonality
 b) Use Minitab to perform a multiplicative decomposition of the time series
of Wal-Mart's sales from the 1990s to 2003 and generate forecasts for
four quarters of 2004. Is a multiplicative decomposition suitable for
the data from Wal-Mart? Explain.

If the multiplicative decomposition is appropriate, in addition to the forecasts for the

69.90
fourth period is 90.58
c) Is the linear trend suitable for these data? If not, could you suggest one?
appropriate trend curve? Adjust your suggested trend curve
and saved the residuals. Calculate the autocorrelations of the residuals. Do the autocorrelations
Do residual correlations suggest a seasonal component? Explain

Adjusted trend equation

Yt = 1165 + 1081.4×t
The multiplicative decomposition is not very suitable since the forecasted values do not
they adjust well to the original data line, but the errors are very high as well.
The simple exponential model is well adjusted with a MAPE error of 11.

The assumptions of normality if they are

they fulfill y also with the one of
homoscedasticity since they follow a pattern
and there is also a random variable since
the data is scattered

As can be seen in the partial autocorrelation and with the assumptions graph, it is possible to
to say that they comply with the seasonal component

Waste
 Autocorrelation function for RESID1
(with significance limits of 5% for the autocorrelations)

1.0

0.8

0.6

0.4

0.2

0.0

-0.2

-0.4

-0.6

-0.8

-1.0

1 5 10 15 20 25 30 35 40 45 50 55
Offset

The smoothing does not provide a reasonable fit for these data as it can be said that the series is
auto correlated and has only 1 lag outside the linear interval.