CHAPTER 7

Forecasting Models
1
9.1 Introduction to Time Series Forecasting
• Forecasting is the process of predicting the future.
• Forecasting is an integral part of almost all business
enterprises.

• Examples
– Manufacturing firms forecast demand for their product, to
schedule manpower and raw material allocation.
– Service organizations forecast customer arrival patterns to
maintain adequate customer service.

2
 Introduction
• More examples
– Security analysts forecast revenues, profits, and
debt ratios, to make investment recommendations.

– Firms consider economic forecasts of indicators

(housing starts, changes in gross national profit)
before deciding on capital investments.

3
 Introduction
• Good forecasts can lead to
– Reduced inventory costs.
– Lower overall personnel costs.
– Increased customer satisfaction.

• The forecasting process can be based on:

– Educated guess.
– Expert opinions.
– Past history of data values, known as a time series.
4
 Components of a Time Series
– Long-term trend
• A time series may be stationary or exhibit trend over time.
• Long-term trend is typically modeled as a linear, quadratic
or exponential function.
– Seasonal variation
• When a repetitive pattern is observed over some time
horizon, the series is said to have seasonal behavior.
• Seasonal effects are usually associated with calendar or
climatic changes.
• Seasonal variation is frequently tied to yearly cycles.

5
 Components of a Time Series
– Cyclical variation
• An upturn or downturn not tied to seasonal variation.
• Usually results from changes in economic conditions.
– Random effects

6
 Components of a Time Series
Time
series
value Linear trend and seasonality time series

Future
Linear trend time series

A stationary time series

Time 7
Steps in the Time Series Forecasting Process

• The goal of a time series forecast is to identify factors

that can be predicted.
• This is a systematic approach involving the following
steps.

– Step 1: Hypothesize a form for the time series model.

– Step 2: Select a forecasting technique.

– Step 3: Prepare a forecast.

8
Steps in the Time Series Forecasting Process

Step 1: Identify components included in the time series

– Collect historical data.

– Graph the data vs. time.
– Hypothesize a form for the time series model.
– Verify this hypothesis statistically.

9
Steps in the Time Series Forecasting Process

• Step 2: Select a Forecasting Technique

– Select a forecasting technique from among several techniques
available. The selection includes
• Determination of input parameter values
• Performance evaluation on past data of each technique

• Step 3: Prepare a Forecast using the selected technique

10
7.2 Stationary Forecasting Models
• In a stationary model the mean value of the time series is
assumed to be constant.
•The values of et
are assumed to be
• The general form of such a model is independent
• The values of et
yt = b 0 + e t are assumed to
have a mean of 0.
Where:
yt = the value of the time series at time period t.
b0 = the unchanged mean value of the time series.
et = a random error term at time period t.
11
 Stationary Forecasting Models
• Checking for trend
• Use Linear Regression if et is normally distributed.
• Use a nonparametric test if et is not normally distributed.
• Checking for seasonality component
• Autocorrelation measures the relationship between the values of the time
series in different periods.
• Lag k autocorrelation measures the correlation between time series
values which are k periods apart.
– Autocorrelation between successive periods indicates a possible trend.
– Lag 7 autocorrelation indicates one week seasonality (daily data).
– Lag 12 autocorrelation indicates 12-month seasonality (monthly data).

• Checking for Cyclical Components

12
 Moving Average Methods

• The last period technique

The forecast for the next period is
the last observed value.

Ft +1 = y t t t
tt+1 t+1 tt+1 t+1

13
 Moving Average Methods

• The moving average method

The forecast is the average of
the last n observations of the
time series.

y t + y t −1 + ... + y t −n+1
Ft +1 =
n t-2
t-2
t-2t-2t-2t-1
t-2
t-1 t t t t t+1
t-1t-1t-1t-1 tt t+1

14
 Moving Average Methods

• The weighted moving average method

– More recent values of the time series
get larger weights than past values when performing
the forecast.

Ft +1 = w1yt + w2yt-1 +w3yt-2 + …+ wnyt-n+1

w1  w2  …  w n
Swi = 1
15
 Moving Average Methods

• Forecasts for Future Time Periods

The forecast for time period t+ 1 is the forecast for

all future time periods:

Ft +k = y t for k = 1, 2, 3,...
This forecast is revised only when new data
becomes available.
16
 YOHO BRAND YO - YOs
Moving Average Methods -

• Galaxy Industries is interested in forecasting weekly

demand for its YoHo brand yo-yos.

• The yo-yo is a mature product. This year demand

pattern is expected to repeat next year.

• To forecast next year demand, the past 52 weeks

demand records were collected.
17
 YOHO BRAND YO - YOs
Moving Average Methods -
• Three forecasting methods were suggested:
– Last period technique - suggested by Amy Chang.
– Four-period moving average - suggested by Bob Gunther.
– Four-period weighted moving average - suggested by Carlos
Gonzalez.

• Management wants to determine:

– If a stationary model can be used.
– What forecast will be obtained using each method?
18
 YOHO BRAND YO YOs- Solution

• Construct the time series plot

W W e ee ek k D Neither
D e em m a an nd d seasonality
W W e ee ek k D D nor
e em m a ancyclical
nd d W W effects
e ee ek k D Dcan
e em m a anbend d observed
W W e ee ek k D D e e m m a a n nd d
11 8 0 04 41 15 5 1 14 4 3 36 65 5 2 27 7 3 35 51 1 4 40 0 2 28 82 2
22 2 23 36 6 1 15 5 4 47 71 1 2 28 8 3 38 88 8 4 41 1 3 39 99 9
33 6 0 03 34 48 8 1 16 6 4 40 02 2 2 29 9 3 33 36 6 4 42 2 3 30 09 9
Demand

44 2 27 72 2 1 17 7 4 42 29 9 3 30 0 4 41 14 4 4 43 3 4 43 35 5
55 4 0 02 8 0 1 18 8 3 37 76 6 3 31 1 3 34 46 6 4 44 4 S e2 9r ie9 s 1
280 299
66 3 39 95 5 1 19 9 3 36 63 3 3 32 2 2 25 52 2 4 45 5 5 52 22 2
200
77 4 43 38 8 2 20 0 5 51 13 3 3 33 3 2 25 56 6 4 46 6 3 37 76 6
88 04 43 31 1 2 21 1 1 19 97 7 3 34 4 3 37 78 8 4 47 7 4 48 83 3
99 4 44 46 6 2 22 2 4 43 38 8 3 35 5 3 39 91 1 4 48 8 4 41 16 6
1

11

16

21

26

31

36

41

46

51
1 10 0 3 35 54 4 2 23 3 5 55 57 7 3 36 6 2 21 17 7 4 49 9 2 24 45 5
1 11 1 5 52 29 9 2 24 4 6 62 25 W5 e e k s 3 37 7 4 42 27 7 5 50 0 3 39 93 3
1 12 2 2 24 41 1 2 25 5 2 26 66 6 3 38 8 2 29 93 3 5 51 1 4 48 82 2
1 13 3 2 26 62 2 2 26 6 5 55 51 1 3 39 9 2 28 88 8 5 52 2 4 48 84 4
19
 Is trend present?

• Run linear regression to test b 1 in the model yt=b 0+b 1t+et

• Excel results
C oe ff. S ta nd . E rr t-S ta t P -va lue Lowe r 9 5 %U ppe r 9 5 %
Inte rce pt 369.27 27.79436 13.2857 5E -18 313.44 425.094
W e e ks 0.3339 0.912641 0.36586 0.71601
0.71601 -1.49919 2.16699

This large P-value indicates

that there is little evidence that trend exists

• Conclusion: A stationary model is appropriate.

20
 Forecast for Week 53
• Last period technique (Amy’s Forecast)
y 53 = y52 = 484 boxes.

– Four-period moving average (Bob’s forecast)

y 53 = (y52 + y51 + y50 + y49) /4 =
(484+482+393+245) / 4 = 401 boxes.
– Four period weighted moving average (Carlo’s forecast)
 =0.4y + 0.3y + 0.2y + 0.1y =
y53 52 51 50 49
0.4(484) + 0.3(482) + 0.2(393) + 0.1(245) = 441.3 boxes.
21
 Forecast for Weeks 54 and 55

• Since the time series is stationary, the forecasts

for weeks 54 and 55 remain as the forecast for
week 53.

• These forecasts will be revised pending

observation of the actual demand in week 53.

22
 The Exponential Smoothing Technique

• This technique is used to forecast stationary time

series.

• All the previous values of historical data affect

the forecast.

23
 The Exponential Smoothing Technique

• For each period create a smoothed value L t of the time

series, that represents all the information known by t.

• The smoothed value Lt is the weighted average of

– The current period’s actual value (with weight of a).
– The forecast value for the current period (with weight of 1-a).

• The smoothed value Lt becomes the forecast for period t+1.

24
The Exponential Smoothing Technique
Define:
Ft+1 = the forecast value for time t+1
yt = the value of the time series at time t
a = smoothing constant

Ft +1 = L t = ay t + (1 − a)Ft

An initial “forecast” is needed to start the process.

25
The Exponential Smoothing Technique –
Generating an initial forecast
– Approach 1:
F2 = L1 = y1
Continue from t=3 with the recursive formula.
– Approach 2:
• Average the initial “ n ” values of the time series.
• Use this average as the forecast for period n + 1 (Fn+1 = y n ).
• Begin using exponential smoothing from that time period
onward [Fn+2 = ay n + (1 − a)Fn+1 = ay n + (1 − a) y n ],
and so on.

26
 The Exponential Smoothing Technique –
Future Forecasts

• Since this technique deals with stationary time series,

the forecasts for future periods does not change.

• Assume N is the number of periods for which data are

available. Then
FN+1 = ayN + (1 – a)FN,
FN+k = FN+1, for k = 2, 3, …
27
 YOHO BRAND YO - YOs
The Exponential Smoothing Technique

• An exponential smoothing forecast is suggested, with a = 0.1.

• An Initial Forecast is created at t=2 by F2 = y1 = 415.

• The recursive formula is used from period 3 onward:

F3 = .1y2 + .9F2 = .1(236) + .9(415) = 397.10
F4 = .1y3 + .9F3 = .1(348) + .9(397.10) = 392.19
and so on, until period 53 is reached (N+1 = 52+1 = 53).
F53 = .1y52 + .9F52 = .1(484) + .9(382.32) = 392.49
F54 = F55 = 392.49 ( = F52) 28
 YOHO BRAND YO - YOs
The Exponential Smoothing Technique
(Excel)

Period Series Forecast

1 415 #N/A
2 236 415
3 348 397.1
45
49 272
280
245 392.19
380.171
382.5884742
50 393 368.8296268
51 482 371.2466641
52 484 382.3219977
53 392.4898
54 392.4898
55 392.4898 29
 YOHO BRAND YO - YOs
The Exponential Smoothing Technique
(Excel)
700
600
500
400
300
200
100
Notice the amount of smoothing
0
0
Included
10
in the smoothed
20 30 40
series
50 60

30
 The Exponential Smoothing Technique
Average age of information
• Relationship between exponential smoothing and simple moving
average
– The two techniques will generate forecasts having the same average age of
information if
2−a
k=
a
– This formula is a useful guide to the appropriate value for a.
• An exponential smoothing forecast “based on large number of periods” should
have a small a.
• A small a provides a lot of smoothing.
• A large a provides a fast response to the recent changes in the
time series and a smaller amount of smoothing.
31
 7.3 Evaluating the performance
of Forecasting Techniques
• Several forecasting methods have been
presented.

• Which one of these forecasting methods gives

the “best” forecast?

32
 Performance Measures

• Generally, to evaluate each forecasting method:

– Select an evaluation measure.

– Calculate the value of the evaluation measure using the forecast

error equation
 t = y t − Ft

– Select the forecast with the smallest value of the evaluation

measure.
33
 Performance Measures –
Sample Example
• Find the forecasts and the errors for each forecasting
technique applied to the following stationary time series.

Time 1 2 3 4 5 6
Time series: 100 110 90 80 105 115
3-Period Moving average: 100 93.33 91.6
Error for the 3-Period MA: - 20 11.67 23.4
3-Period Weighted MA(.5, .3, .2) 98 89 85.5
Error for the 3-Period WMA - 18 16 29.5

34
 Performance Measures

S( )2t Sn|t| yt

MSE =
n MAPE = n

MAD =
S | t| LAD = max | t|
n

35
 Performance Measures –
MSE for the Sample Example

MSE for the moving average technique:

MSE =
S (t)2 (-20) +(11.67) +(23.4)
=
2 2 2
= 361.24
n 3

MSE for the weighted moving average technique:

MSE =
S (t)2 (-18) + (16) + (29.5)
=
2 2 2
= 483.4
n 3
36
 Performance Measures –
MAD for the Sample Example

MAD for the moving average technique:

MAD =
S | t| = |-20| + |11.67| + |23.4|
= 18.35
n 3

MAD for the weighted moving average technique:

MAD =
S | t| = |-18| + |116| + |29.5| = 21.17
n 3
37
 Performance Measures –
MAPE for the Sample Example

MAPE for the moving average technique:

S | t| |-20|/80 + |11.67|/105+ |23.4|/115

MAPE= = = .188
n 3

MAPE for the weighted moving average technique:

S | t| |-18|/80 + |16|/105 + |29.5|/115

MAPE= = = .211
n 3
38
 Performance Measures –
LAD for the Sample Example

LAD for the moving average technique:

LAD= max | t| = max {|-18|, |16|, |29.5|} = 29.5

LAD for the weighted moving average technique:

LAD= max | t|= |-20|, |11.67|, |23.4| = 23.4

39
Performance Measures –
YOHO BRAND YO - YOs
=B4
Drag to Cell C56

=E5/B
5
=D5^
2
=ABS(D5 Highlight
) Cells D5:G5
=B5-C5
and Drag to
Cells
D55:G55
40
Performance Measures –
YOHO BRAND YO - YOs

=AVERAGE(B4:B7)
Forecast begins at period 5. Drag to Cell C56

=E8/B
8
=D8^
2 Highlight
=ABS(D8
) Cells D8:G8
=B8-C8 and Drag to
=C56 Cells
Drag to C58 D55:G5541
 Performance Measures –
Selecting Model Parameters
• Use the performance measures to select a good set of
values for each model parameter.
– For the moving average:
• the number of periods (n).
– For the weighted moving average:
• The number of periods (n),
• The weights (Wi).
– For the exponential smoothing:
• The exponential smoothing factor (a).
• Excel Solver can be used to determine the values of the
model parameters.
42
Weights for the Weighted Moving Average
Minimize MSE using Solver-

Minimize Cell containing MSE

Weights

Weights Are
Nonincreasin
g
Total of Weights
Sum to 1

43
 Selecting Forecasting Technique

• Key issues considered when determining the

technique to be used in forecasting stationary time-
series.
– The degree of autocorrelation.

– The possibility of future shifts in time series values.

– The desired responsiveness of the forecasting technique.

– The amount of data required to calculate the forecast.

44
 7.4 Time Series with Linear Trend

• If we suspect trend, we should assess whether the

trend is linear or nonlinear.
• Here we assume only linear trend.
yt = b 0 + b 1t + et

• Forecasting methods
– Linear regression forecast.
– Holt’s Linear Exponential Smoothing Technique.
45
 The Linear Regression Approach

• Construct the regression equation based on the

historical data available.
• The independent variable is “time”.
The dependent variable is the “time-series value”.
• Forecasts should not expand to periods far into the
future.

46
 Holt’s Technique –
A qualitative demonstration

Y3 F4
+ L3 − L2
L t = ay t + (1 − a)Ft
T2 T3
L 3+
++ Tt = (L t − L t −1 ) + (1 −  )Tt −1
++
F3+
+T2
L+2

2 3 4
47
 The Holt’s Technique
• The Holt’s Linear Exponential Smoothing Technique.
– Adjust the Level Lt , and the Trend Tt in each period:
Level: L t = ay t + (1 − a )Ft
Trend: Tt = (L t − L t −1 ) + (1 −  )Tt −1
Initial values: L 2 = y 2 and T2 = y 2 − y 1
a = smoothing constant for the time series level.
 = smoothing constant for the time series trend.
L t = estimate of the time series for time t as of time t.
Tt = estimate of the time series trend for time t as of time t.
yt = value of the time series at time t.
Ft = forecast of the value of the time series for time t calculated at a
period prior to time t.
48
 Future Forecasts
• Forecasting k periods into the future
– By linear regression

Ft +k = b0 + b1(t + k)

– By the Holt’s linear exponential smoothing technique

Ft +k = L t + kTt

49
 American Family Products Corp.
• Standard and Poor’s (S&P) is a bond rating firm.

• It is conducting an analysis of American Family

Products Corp. (AFP).
• The forecast of year-end current assets is
required for years 11 and 12, based on data over
the previous 10 years.

50
 American Family Products Corp.
• The company’s assets have been increasing
at a relatively constant rate over time.
• Data
Year-end current assets
Year Current Assets
1 1990 (Million)
2 2280
3 2328
4 2635
5 3249
6 3310
7 3256
8 3533
9 3826
10 4119 51
 AFP -SOLUTION
Forecasting with the Linear Regression Model
A s s e ts
4500
4000
3500
3000
2500
2000
1500
1000
500
0
Y ear 1 2 3 S1
4 5 6 7 8 9
Ye ar 10

52
 AFP -SOLUTION
Forecasting with the Linear Regression Model
R egression S tatistics
M ultiple R 0.980225251
R S quare 0.960841543 The Regression Equation
A djusted R S quare
0.955946736 y t = 1788.2 + 229.89 t
S tandard E rror 149.0358255
O bservations 10

ANOVA
df SS MS F S ignificance F
R egression 1 4 3 6 0 1 1 0 .9 8 2 4 3 6 0 1 1 0 .9 8 2 196.2981421 6.53249E -07
R esidual 8 177693.4182 22211.67727
T otal 9 4 5 3 7 8 0 4 .4

C oefficients S tandard E rror t S tat P -value Low er 95% U pper 95%

Intercept 1788.2 101.8108511 17.56394315 1.12773E -07 1553.423605 2022.976395
Y ear 229.8909091 16.40830435 14.01064389 6.53249E -07 192.0532669 267.7285513
53
 AFP -SOLUTION
Forecasting with the Linear Regression Model

=$B$31+$B$32*A2
Drag to cells C3:C13

Forecasts for Years

11 and 12

54
 AFP -SOLUTION
Forecasting using the Holt’s technique
Demonstration of the calculation procedure. Year Current Assets
with a = 0.1 and  = 0.2 1 1990
Year 2: y2 = 2280 2 2280
3 2328
L2 =y2 L2 = 2280.00
4 2635
T2=y2-y1 T2 = 2280 - 1990 = 290 5 3249
F3=L2+T2 F3= 2280 + 290 = 2570.00 ………………………
Year 3: y3 = 2328 ………………………

L t = ay t + (1 − a)Ft L3 = (.1)(2328) + (1 - 0.1)(2570.00) = 2545.80

Tt = (L t − L t −1 ) + (1 −  )Tt −1 T3 = (.2)(2545.80-2280)) + (1 - 0.2)(290.00) = 285.16
Ft+1=Lt+Tt F4= 2545.80 + 285.16 = 2830.96
55
 AFP -SOLUTION
Forecasting using the Holt’s technique (Excel)
INPUTS OUTPUTS
Number of Periods of Data Collected = 10 Period Forecast =
Smoothing Constant (alpha) = 0.1
Smoothing Constant (gamma) = 0.2 MSE = 72994.37 MAPE = 7.752572
Initial Forecast Value (Level) = 2280 MAD = 250.3462 LAD = 411.0682
Initial Forecast Value (Trend) = 290

OUTPUTS
Period 11 12 13 14 15 16 17 18 19
Forecast 4593.378 4849.579 5105.779 5361.98 5618.18 5874.38 6130.581 6386.781 6642.982

Forecast Forecast Absolute Error Absolute

Period Value Level Trend Forecast Error Error Squared % Error
1 1990
2 2280 2280 290
3 2328 2545.8 285.16 2570 -242 242 58564 0.103952
4 2635 2811.364 281.2408 2830.96 -195.96 195.96 38400.32 0.074368
5 3249 3108.244 284.3687 3092.605 156.3952 156.3952 24459.46 0.048136
6 3310 3384.352 282.7164 3392.613 -82.613 82.61302 6824.912 0.024959
7 3256 3625.961 274.4951 3667.068 -411.068 411.0682 168977 0.126249
8 3533 3863.711 267.146 3900.456 -367.456 367.4564 135024.2 0.104007
9 3826 4100.371 261.0488 4130.857 -304.857 304.8567 92937.63 0.07968
10 4119 4337.178 256.2004 4361.42 -242.42 242.4199 58767.4 0.058854
11 4134.04 164.3329 4593.378
56
7.5 Time Series with Trend, Seasonality,
and Cyclical Variation
• Many time series exhibit seasonal and cyclical variation
along with trend.
• Seasonality and cyclical variation arise due to calendar,
climate, or economic factors.
Time series value
• Two models are considered: Trend component
– Additive model Cyclical component
yt = Tt + Ct + St + et
Random error
– Multiplicative model Seasonal component
yt = Tt Ct Stet 57
 The Classical Decomposition
• This technique can be used to develop an additive or
multiplicative model.
• The time series is first decomposed to its components
(trend, seasonality, cyclical variation).
• After these components have been determined, the
series is re-composed by
– adding the components - in the additive model
– multiplying the components - in the multiplicative model.

58
The Classical Decomposition- Procedure (1)

• Smooth the time series to remove • Calculate moving averages.

random effects and seasonality.

• Determine “period factors” to • Calculate the ratio yt/MAt.

isolate the (seasonal)•(error)
factor.

• Determine the “unadjusted • Average all the yt/MAt that

seasonal factors” to eliminate the correspond to the same season.
random component from the period
factors

59
The Classical Decomposition- Procedure (2)

• Determine the “adjusted seasonal Calculate:

factors”. [Unadjusted seasonal factor]
[Average seasonal factor]
• Determine “Deseasonalized data Calculate:
values”. yt
[Adjusted seasonal factors]t

• Determine a deseasonalized trend Use linear regression on the

forecast. deseasonalized time series.

• Determine an “adjusted seasonal Calculate:

forecast”. (yt/Mat) •[Adjusted seasonal forecast].
60
CANADIAN FACULTY ASSOCIATION (CFA)

• The CFA is the exclusive bargaining agent for the

state-supported Canadian college faculty.
• Membership in the organization has grown over the
years, but in the summer months there was always
a decline.
• To prepare the budget for the 2001 fiscal year, a
forecast of the average quarterly membership
covering the year 2001 is required.
61
 CFA - Solution

• Membership records from 1997 through 2000 were

collected and graphed.

62
 CFA - Solution
AVERAGE
YEAR PERIOD QUARTER MEMBERSHIP
1997 1 1 7130
2 2 6940
3 3 7354
4 4 7556
1998 5 1 7673 Average Dues Paying Members (payroll deduction)
6 2 7332
7 3 7662 9000

8 4 7809 The graph exhibits long term trend

1999 9 1 7872 8500

10 2 7551 The graph exhibits seasonality pattern

11 3 7989 8000

12 4 8143
2000 13 1 8167 7500

14 2 7902
15 3 8268 7000

16 4 8436
6500

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

1997 1998 P er i od
1999 2000
63
 Classical Decomposition – step 1:
Isolating Trend and Cyclical Components

• Smooth the time series to

remove random effects and Calculate moving averages.
seasonality.
Average membership for the first 4 periods First moving average is centered at
= [7130+6940+7354+7556]/4 = 7245.01 quarter (1+4)/ 2 = 2.5

Average membership for periods [2, 5] Second moving average is centered at

= [6940+7354+7556+7673]/4 = 7380.75 quarter (2+5)/ 2 = 3.5

Centered moving average of the first

two moving averages is Centered location is t = 3
[7245.01 + 7380.75]/2 = 7312.875 64
 Classical Decomposition – step 2
[Seasonal•Random Error] Factors
• Determine “period factors” to
isolate the Calculate the ratio yt/MAt.
(Seasonal)•(Random error)
factor.
The Centered Moving Average only
represents TtCt. The Seasonal•Random
error factors are represented by
Stet = yt/TtC t

Example: In period 7 (3 rd quarter of

1998):
S7e7=7662/7643.875 = 1.00237 65
 Classical Decomposition – step 3
Unadjusted Seasonal Factors
• Determine the “unadjusted Average all the yt/MAt that
seasonal factors” to eliminate correspond to the same
the random component from season.
the period factors

Averaging the Seasonal•Random Error Factors eliminates the random factor

from Ste t . This leaves us with the seasonality component only for each season.

Example: Unadjusted Seasonal Factor for the third quarter.

S3 = {S3,97 e3,97 + S3,98 e3,98 + S3,99 e3,99}/3 = {1.0056+1.0024+1.0079}/3 = 1.0053

66
 Classical Decomposition – step 4
Adjusted Seasonal Factors
• Determine the “adjusted Calculate:
seasonal factors”. [Unadjusted seasonal factor]
[Average seasonal factor]
Without seasonality the seasonal factors for each season should be equal to 1.
Thus, the sum of all seasonal factors would be equal to 4.
The adjustment of the unadjusted seasonal factors maintains the sum of 4.
Example: The average seasonal factor is
(1.0149+.9658+1.00533+1.01624)/4=1.00057.
Adjusted Seasonal
factors The adjusted seasonal factor for the 3rd quarter:
Quarter 1 1.014325 S3/Average seasonal factor = 1.00053/1.00057 = 1.00472
Quarter 2 0.965252 Excel’s exact value=1.004759).
Quareter 3 1.004759
Quarter 4 1.015663 67
 Classical Decomposition – step 5
The Deseasonalized Time Series
• Determine “Deseasonalized Calculate:
data values”. yt
[Adjusted seasonal factors]t

Deseasonalize the Time Series by .yt/(Adjusted S)t = TtCtet

Example:
Deseasonalized series value for the 2nd quarter, 1998 =
y6/[Adjusted S2] = 7332/0.9652 = 7595.94

68
 Classical Decomposition – step 5
The Time series trend
Deseasonalized Time Series
8400
Average Membership

Seasonality has been substantially

(payroll deduction)

8200

8000
Reduced. This graph represents TtCtet.
7800

7600

7400
Average Dues Paying Members (payroll deduction)
7200
9000

7000 8500

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

8000

Period
7500

7000

6500

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

P er i od 69
Classical Decomposition – step 6
The Time series trend Component

• Determine a deseasonalized
trend forecast. Use linear regression on the
deseasonalized time series.

Trend factor: Tt = 7069.6677 + 78.4046t

70
Classical Decomposition – step 7
The forecast

• Assuming no cyclic effects the forecast

becomes:
F(quarter i, time N+k) = TN+k•(Adjusted Si)

Trend factor: T17 = 7069.6677 + 78.4046(17) = 8402

Forecast(Q1, t=17) = (8402)(1.01433) = 8523

71
 Classical Decomposition – step 7
The forecast
• Assuming cyclic effects, create a series
of the cyclic component, as follows:
• Deseasonalized time series TCe
= t t t = Ctet.
Trend component (from the regression) Tt

• Smooth out the error component using moving averages to

isolate Ct.

• Perform the forecast:

F(quarter i, time N+k) = TN+k CN (Adjusted Si)
• •

72
Classical Decomposition Template

73
 The additive model –
The Multiple Regression Approach
For a time series with trend and seasonality:

Yt = Tt +St + Rt, which translates to

Yt = b0 + b1t + b2S1 + … +bkSk + et

The trend element The seasonality element

74
 Troy’s Mobil Station
• Troy owns a gas station that experience seasonal
variation in sales.
• In addition, due to a steady increase in population
Troy feels that average sales are increasing
generally.

75
 Troy’s Mobil Station

• Data
Gasoline Sales Over Five Year Period
5000

4800
Gasoline Sales
Average Daily

4600

4400
(gallons)

4200

4000

3800

3600

3400

3200

3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

F W Spg Smr P e r i od
76
 Troy’s Mobil Station –
Multiple Regression input data
Trend Seasonal variables
variable
Quarterly Input Data
Sales t X1 X2 X3
3497 1 Fall
1 0 0
3484 2 0 Winter
1 0
Year 1
3553 3 0 0 Spring
1
3837 4 Not0Fall Not Winter
0 Not 0
Spring
3726 5 1 0 0
Year 2
3589 6 0 1 0

77
Troy’s Mobil Station Template

78
 Troy’s Mobil Station –
Multiple regression (Excel output)

Extremely good fit

All the variable

are linearly related Extremely useful
to sales.

79
 Troy’s Mobil Station –
Multiple regression (Graphical interpretation)

-155.00
• •

-248.27

-155.00 -322.93 •
•
•
Fall Winter Spring Summer

80
 Troy’s Mobil Station –
Performing the forecast
• The forecasting additive model is:
Ft = 3610.625 + 58.33t – 155F – 323W – 248.27S

• Forecasts for year 5 are produced as follows:

• F(Year 5, Fall) = 3610.625+58.33(21) – 155(1) – 323(0) – 248.27(0)
• F(Year 5, Winter) = 3610.625+58.33(22) – 155(0) – 323(1) – 248.27(0)
• F(Year 5, Spring) = 3610.625+58.33(23) – 155(0) – 323(0) – 248.27(1)
• F(Year 5, Summer) = 3610.625+58.33(24) – 155(0) – 323(0) – 248.27(0)

81
Copyright © 2002 John Wiley & Sons, Inc. All rights
reserved. Reproduction or translation of this work beyond
that named in Section 117 of the United States Copyright
Act without the express written consent of the copyright
owner is unlawful. Requests for further information
should be addressed to the Permissions Department, John
Wiley & Sons, Inc. Adopters of the textbook are granted
permission to make back-up copies for their own use
only, to make copies for distribution to students of the
course the textbook is used in, and to modify this material
to best suit their instructional needs. Under no
circumstances can copies be made for resale. The
Publisher assumes no responsibility for errors, omissions,
or damages, caused by the use of these programs or from
the use of the information contained herein.
82