STATGRAPHICS – Rev.

7/20/2009

Seasonal Decomposition

Summary
The Seasonal Decomposition procedure divides a time series into three components:

1. trend-cycle
2. seasonality
3. irregular

Each component may be separately plotted or saved. In addition, the decomposition can
be used to create a seasonally adjusted version of the original time series. Seasonal
subseries and annual subseries plot may also be created.

Sample StatFolio: tsdecomp.sgp

Sample Data:
The file golden gate.sgd contains monthly traffic volumes on the Golden Gate Bridge in
San Francisco for a period of n = 168 months from January, 1968 through December,
1981. The table below shows a partial list of the data from that file:

Month Traffic
1/68 73.637
2/68 77.136
3/68 81.481
4/68 84.127
5/68 84.562
6/68 91.959
7/68 94.174
8/68 96.087
9/68 88.952
10/68 83.479
11/68 80.814
12/68 77.466
1/69 75.225
… …

The data were obtained from a publication of the Golden Gate Bridge.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 1

 STATGRAPHICS – Rev. 7/20/2009

Data Input
The data input dialog box requests the name of the column containing the time series
data:

 Data: numeric column containing n equally spaced numeric observations.

 Time indices: time, date or other index associated with each observation. Each value
in this column must be unique and arranged in ascending order.

 Sampling Interval: If time indices are not provided, this defines the interval between
successive observations. For example, the data from the Golden Gate Bridge were
collected once every month, beginning in 01/68.

 Seasonality: the length of seasonality s, or number of observations in a full cycle of

the seasonal pattern. For example, monthly data such as traffic on the Golden Gate
Bridge have a seasonality of s = 12. Hourly data that repeat every day have a
seasonality of s = 24.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 2

 STATGRAPHICS – Rev. 7/20/2009

 Trading Days Adjustment: a numeric variable with n observations used to

normalize the original observations, such as the number of working days in a month.
The observations in the Data column will be divided by these values before being
plotted or analyzed.

 Select: subset selection.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 3

 STATGRAPHICS – Rev. 7/20/2009

Statistical Model
In analyzing time series data, it is common to view the data as consisting of several
components:

1. Trend (T) – a general long-term pattern observed over the entire data set. For
example, many economic time series tend to show an increasing trend when
viewed over many years.

2. Cycle (C) – cyclical variations around the trend line. Unlike seasonal effects,
these cycles do not have a fixed frequency. General up and downs of the general
world economy is a typical example.

3. Seasonality (S) – cyclical variations with a fixed frequency, such as yearly cycles
in the sales of lawnmowers. Seasonal effects repeat on a regular and predictable
basis.

4. Random or Irregular (R) – the residual component left behind after the other three
components are accounted for.

There are two basic models upon which a decomposition of a time series into its
component parts may be based: a multiplicative model and an additive model. The
multiplicative model assumes that the data at time t may be represented as the product of
the four components according to:

Yt  Tt C t S t Rt (1)

The additive model assumes that the components add:

Yt  Tt  C t  S t  Rt (2)

The goal of the Seasonal Decomposition procedure is to divide an observed time series
into its component parts. In particular, the procedure derives:

1. Seasonal indices representing the effect of each season. Knowing the effect of
different seasons is often quite important.

2. An estimated of the combined trend-cycle component. No attempt is made to

separate these two components, however, since both represent relatively long-
term effects.

3. The residuals (irregular component).

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 4

 STATGRAPHICS – Rev. 7/20/2009

Analysis Summary
The Analysis Summary displays the number of observations in the time series, the length
of seasonality, and the decomposition method selected.

Seasonal Decomposition - Traffic

Data variable: Traffic (Golden Gate Bridge Traffic Volume)

Number of observations = 168

Start index = 1/68
Sampling interval = 1.0 month(s)
Length of seasonality = 12

Seasonal Decomposition
Method: Multiplicative

Note: a limited amount of missing data is permitted, providing there are not too many
missing values close together. Missing values are replaced by interpolated values
according to the method outlined in the Calculations section of the Time Series –
Descriptive Methods documentation.

Analysis Options
Analysis Options permits the data to be transformed before being plotted or analyzed:

 Math: transforms the data by performing the indicated mathematical operation.

 Inflation: adjusts the data for inflation using the specified inflation rate

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 5

 STATGRAPHICS – Rev. 7/20/2009

If more than one transformation is requested, they are applied in the following order:

1. trading day adjustment

2. inflation adjustment
3. math adjustment

For a detailed discussion of the transformation options, refer to the Time Series –
Descriptive Methods documentation.

Data Table
The Data Table displays the results of the decomposition:

Data Table for Traffic

Seasonal decomposition method: Multiplicative

Seasonally
Period Data Trend-Cycle Seasonality Irregular Adjusted
1/68 73.637 81.0636
2/68 77.136 82.2507
3/68 81.481 84.2308
4/68 84.127 84.6577
5/68 84.562 84.4944
6/68 91.959 87.0608
7/68 94.174 84.5557 111.375 103.054 87.1383
8/68 96.087 84.7169 113.421 102.394 86.7454
9/68 88.952 84.9508 104.71 100.136 85.0664
10/68 83.479 85.1548 98.032 98.5183 83.8931
11/68 80.814 85.342 94.6943 98.1288 83.7451
12/68 77.466 85.5072 90.5959 96.2271 82.2811
1/69 75.225 85.6004 87.8793 96.7423 82.8118
2/69 79.418 85.7466 92.6194 98.7608 84.684
3/69 84.813 86.0028 98.6166 101.945 87.6753
4/69 85.691 86.3519 99.2347 99.8607 86.2316
5/69 87.49 86.7945 100.801 100.721 87.42
6/69 92.995 87.3052 106.517 100.844 88.0417
7/69 95.375 87.7297 108.715 100.593 88.2496
8/69 98.396 88.1522 111.621 100.769 88.83
9/69 92.791 88.5946 104.737 100.161 88.7377

Included in the table are:

 Data: the original time series Yt, including any replacement values that have been
calculated for missing data.

 Trend-Cycle: an estimate of the combined trend-cycle component (TtCt for a

multiplicative decomposition and Tt + Ct for an additive decomposition).

 Seasonality: the estimated seasonal component St.

 Irregular: the estimated residual or irregular component Rt.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 6

 STATGRAPHICS – Rev. 7/20/2009

 Seasonally Adjusted Data: the original data with only the seasonality removed.

Pane Options

 Method: the type of decomposition performed.

Trend-Cycle Plot
The Trend-Cycle plot shows the estimated trend-cycle component.

Trend-Cycle Component Plot for Traffic

113
data
trend-cycle
103
Traffic

93

83

73
1/68 1/71 1/74 1/77 1/80 1/83

The trend-cycle component is estimated by smoothing the time series data using a simple
moving average with span k equal to the length of seasonality s. In the traffic data shown
above, the general pattern is upward, although there were two major shocks to the system
that had a major impact on the basic trend.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 7

 STATGRAPHICS – Rev. 7/20/2009

Seasonal Indices
Once the trend-cycle has been estimated, it can be removed from the data. For a
multiplicative model, this is done by dividing the original data by the estimated
component (called a ‘ratio-to-moving average”), leaving:

Y
Sˆ t Rˆ t  t (3)
TˆCˆ t

For an additive model, the trend-cycle is subtracted from the original data, leaving:

Sˆ t Rˆ t  Yt  TˆCˆ t (4)

The resulting estimates of the seasonal-irregular component are then averaged across all
observations within each season to remove the irregular component, resulting in an
estimate of the seasonal component. The seasonal components are then adjusted so that
an average season has a value of 1.0 if using the multiplicative method and 0 if using the
additive method.

The Seasonal Indices table displays the results:

Seasonal Indices for Traffic

Seasonal decomposition method: Multiplicative

Season Index
1 90.8385
2 93.7815
3 96.7354
4 99.3731
5 100.08
6 105.626
7 108.074
8 110.769
9 104.568
10 99.5064
11 96.5
12 94.148

This table shows the estimated seasonal component St in the traditional format of a
percentage, such that an average season would have an index equal to 100. For example,
the traffic index in August is approximately 110.8, meaning that traffic across the Golden
Gate Bridge in August is 10.8% higher than average.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 8

 STATGRAPHICS – Rev. 7/20/2009

Seasonal Indices Plot

The Seasonal Indices plot displays the estimated indices:

Seasonal Index Plot for Traffic

115

110
seasonal index

105

100

95

90
0 3 6 9 12 15
season

Irregular Component
The Irregular Component plot displays the residual or irregular component Rt at each
time period:

Irregular Component Plot for Traffic

105

100
irregular

95

90

85
1/68 1/71 1/74 1/77 1/80 1/83

For a multiplicative model, the estimated irregular component is obtained by dividing the
observations by the estimated trend-cycle and seasonal components:

Yt
Rˆ t  (5)
TˆCˆ Sˆ
t t

and then normalized so that the average residual equal 1.0 (corresponding to an index of
100). For an additive model, the irregular component is obtained by subtraction:

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 9

 STATGRAPHICS – Rev. 7/20/2009

Rˆ t  Yt  TˆCˆ t  Sˆ t (6)

and then normalized so that the average residual equals 0.

In the sample data, note the large residual in March of 1974, equaling approximately
86.4. This indicates that traffic in that month was 13.6% below what would have been
expected, given the estimated trend-cycle and seasonal effects.

Seasonally Adjusted Data

Once all of the components have been estimated, the original time series can be
seasonally adjusted by removing from it only the seasonal effects, leaving both the trend-
cycle and irregular components. For a multiplicative decomposition, the seasonally
adjusted data is given by

Yt
Yt adj  (7)
Sˆ
t

For an additive decomposition, the seasonally adjusted data is given by

Yt adj  Yt  Sˆ t (8)

The Seasonally Adjusted Data plot displays the results:

Seasonally Adjusted Data Plot for Traffic

106
seasonally adjusted

101
96
91
86
81
76
1/68 1/71 1/74 1/77 1/80 1/83

The effect of the gasoline shortages on the basic trend line is now quite apparent.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 10

 STATGRAPHICS – Rev. 7/20/2009

Seasonal Subseries Plot

An alternative way to plot seasonal data is through a Seasonal Subseries Plot:

Seasonal Subseries Plot for Traffic

113

103
Traffic

93

83

73
0 2 4 6 8 10 12 14
Season

This plot is constructed as follows:

1. The observations corresponding to each season are collected and horizontal lines
drawn at the average for the season.

2. Vertical lines are drawn from each observation to the average of the season it
corresponds to.

In such a plot, one can see all of the time series components:

(i) The seasonal pattern is observable by looking at the differences

between the averages of each season.
(ii) The trend-cycle is observable by looking at patterns within each
season.
(iii) The irregular component is observable by looking at the length of the
vertical lines contrasted with the trend within each season.

Pane Options

 Vertical Lines – draw a line from each observation to the average for its season.
 Connected Scatterplots – draw the data within each season as a connected X-Y plot.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 11

 STATGRAPHICS – Rev. 7/20/2009

Annual Subseries Plot

The Annual Subseries Plot displays each cycle as a separate plot:

Annual Subseries Plot for Traffic Cycle

1968
113 1969
1970
103 1971
1972
Traffic

1973
93 1974
1975
83 1976
1977
1978
73 1979
0 3 6 9 12 15 1980
Season 1981

Pane Options

 Cumulative: if checked, the value plotted on the vertical axis is the cumulative sum
of the observations during the cycle.

Example – Cumulative Plot

Annual Subseries Plot for Traffic Cycle

1968
1500 1969
1970
1200 1971
1972
Traffic

900 1973
1974
600 1975
1976
300 1977
1978
0 1979
0 3 6 9 12 15 1980
Season 1981

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 12

 STATGRAPHICS – Rev. 7/20/2009

Save Results
The following results may be saved to the datasheet:

1. Data – the original time series data, after replacement of any missing values.

2. Trend-cycle – the estimated trend-cycle component.

3. Seasonal indices – the estimated seasonal indices.

4. Irregular – the irregular component.

5. Seasonally adjusted data – the seasonally adjusted data.

Note: if a multiplicative model was used, the Seasonal indices and Irregular components
will be normalized so that the average value equals 100. For an additive model, the
Seasonal indices will be normalized so that the average value equals 0.

2009 by StatPoint Technologies, Inc. Seasonal Decomposition - 13