Time series Decomposition

Additive Model

Farideh Dehkordi-Vakil
Classical Decomposition
 Additive Decomposition
 We assume that the time series is additive. A
classical decomposition can be carried out
using the following steps.
 Step 1: The trend cycle is computed using a centered
MA of order k.
 Step2: The detrended series is computed by
subtracting the trend-cycle component from the data
Yt  Tt  St  Et
Classical Decomposition
 Additive Decomposition
 Step3: In classical decomposition we assume the
seasonal component is constant from year to year.
So we the average of the detrended value for a given
month (for monthly data) and given quarter (for
quarterly data) will be the seasonal index for the
corresponding month or quarter.
 Step4: the irregular series Et is computed by simply
subtracting the estimated seasonality, and trend-
cycle from the original data.
The Time-Series Decomposition Forecast
 We have seen that using multiplicative model, a
time series data can be decomposed into the
product of four components:
Yt  Tt  St  Ct  Et
 Y= The series to be forecast.
 T= The long–term trend based on deseasonalized data.
 It is often called the centered moving-average trend (CMAT)
since the deseasonalized data are centered moving averages
(CMA) of the original Y values.
 S = Seasonal indexes (SI).
 These are normalize average of seasonal factors that are
determined as the ratio of each period’s actual value y to the
deseaonalized value (CMA) for that period.
The Time-Series Decomposition Forecast
 C = The cycle component.
 The cycle factor (CF) is the ratio of CMA to CMAT
and represents the gradual wavelike movements in
the series around the trend line.
 E = The irregular component.
 This is assumed equal to 1 unless the forecasters has
reason to believe a shock may take place, in which
case I could be different from 1 for all or part of the
forecast period.
The Time-Series Decomposition Forecast
 We know how to isolate and measure these
components.
 To prepare a forecast based on the time
series decomposition model, we must
reassemble the components.
 The forecast for Y (FY) is:
FY  (CMAT)(SI) (CF)(E)
Example:Private Housing Start
Example:Private Housing Start
Example:Private Housing Start
 The series appears quite
volatile
 The sharp increases and
decreases appears to
follow a reasonably
regular manner, which
may reflect a seasonal
component.
 There also appears to be
some long-term wavelike
movement to the data as
well as a slight positive
trend.
Example:Private Housing Start
Example:Private Housing Start
 The centered moving
average series, shown by
the solid line, is much
smoother than the original
series of private housing
starts data (dashed line)
because the seasonal
pattern and the irregular or
random fluctuation in the
data are removed by the
process of calculating the
centered moving average.
Example:Private Housing Start
 The long term trend in private
housing starts is shown by the
straight dotted line
(PHSCMAT).
 The dashed line is the raw data
(PHS), while the wavelike solid
line is the deseasonalized data
(PHSCMA).
 The long-term trend is positive.
The equation for the trend line
is:
PHSCMAT  237.51  0.313(Time)
Example:Private Housing Start
 The cyclical factor is
the ratio of the
centered moving
average to the long-
term trend in the data.
 As this plot shows, the
cycle factor moves
slowly around the base
line (1.0) with little
regularity
Example:Private Housing Start
Example:Private Housing Start
 The actual values for
private housing starts
are shown by the
dashed line, and the
forecast values based
on the time- series
decomposition model
are shown by the solid
line.
The Time-Series Decomposition Forecast

 Because the time series decomposition

models do not involve a lot of mathematics
or statistics, they are relatively easy to
explain to the end user.
 This is a major advantage because if the end
user has an appreciation of how the forecast
was developed, he or she may have more
confidence in its use for decision making.