Time Series Analysis and Forecasting

Statistics
for Management & Economics 10e

Identify • Compute • Interpret
GERALD KELLER • WILFRID LAURIER UNIVERSITY
© 2015 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part, except for use as permitted in a
license distributed with a certain product or service or otherwise on a password-protected website for classroom use.
Chapter 20
Time Series Analysis and

Forecasting
license distributed with a certain product or service or otherwise on a password-protected website for classroom use. 20.2
Time Series Analysis…
A variable measured over time (in sequential order) is called
a time series. From this data, we analyze it to detect patterns
that will enable us to forecast future values of the variable.
As you might expect, this technique has wide application:

u Governments want to know future values of interest rates,
unemployment rates and percentage increases in the cost of living.
v Housing industry economists must forecast mortgage interest rates,
demand for housing, and the cost of building materials.
w Many companies attempt to predict the demand for their product
and their share of the market.
x Universities and colleges often try to forecast the number of
students who will be applying for acceptance at post-secondary-school
institutions.
…and so on!
Time Series Components…
A time series can consist of four different components:
u Long-term trend
v Cyclical variation
variable of interest
w Seasonal variation
x Random variation
A trend is a long term

relatively smooth pattern or
direction, that persists usually
for more than one year
time
u Long-term trend
x Random variation
A cycle is a wavelike pattern

describing a long term behavior
(for more than one year).
time
Cyclical patterns that are consistent
and predictable are quite rare; hence,
we will ignore this type of variation
u Long-term trend
x Random variation
The seasonal component of

the time series exhibits a short
term (less than one year)
calendar repetitive behavior.
time
u Long-term trend
x Random variation
Random variation comprises

the irregular unpredictable
changes in the time series. It
tends to hide the other (more
predictable) components.
time
One of our objectives will be to
remove random variation…
Smoothing Techniques…
If we can determine which components actually exist in a
time series, we can develop better forecasts.
We can reduce random variation by smoothing the time

series.
To methods to smooth the data are:

 moving averages, and
 exponential smoothing.
Moving Averages…
A moving average for a time period is the arithmetic mean
of the values in that time period and those close to it. This is
what we hear something like “three year moving average”.
time
1 2 3 4 5 …
Example 20.1: period
sales
[Xm20-01] 39 37 61 58 18 …
(,000s)
and so on…
Example 20.1… COMPUTE
Again, calculating manually can be tedious and error prone.

In Excel: Data > Data Analysis > Moving Average
note how the moving average is

“smoother” than the raw data…
Moving Average… INTERPRET
The averaging process removes some of the random variation. If we

use a 5-quarter moving average, removes even more variation and
our line is smoother, but now we’ve lost the seasonality that appears in
the 3-quarter moving average.
Centered Moving Average…
We’ve considered moving averages for odd numbers of time
periods:
3-period:
5-period:
What happens when we use even numbers of periods to

calculate moving averages?
With an even number of observations included in the
moving average, the average is placed between the two
periods in the middle.
To place the moving average in an actual time period, we

need to center it.
Two consecutive moving averages are centered by taking

their average, and placing it in the middle between them.
Consider this 6-period time series:
15, 27, 20, 14, 25, 11
Can we calculate its four-period moving average?
Exponential Smoothing…
There are two drawbacks with the moving average method
of smoothing:
 No moving averages for the first and last sets of
time periods.
 The moving average “forgets” most of the previous
time-series values (i.e. only looks at those around it).
Exponential smoothing addresses these issues…
Exponentially Smoothed Time Series…
An exponentially smoothed time series is one that’s given by
St = wyt + (1 – w)St-1 (for t ≥ 2)
where:
St = Exponentially smoothed time series at time t
yt = Time series at time period t (our original data)
St-1 = Exponentially smoothed time series at time t–1

w = Smoothing constant, where 0 ≤ w ≤ 1
In general:
We can calculate these values manually…

S1 = y1
St = wyt + (1 – w)St-1
Excel > Data > Data Analysis > Exponential Smoothing

Xm20-02
1–w
Example 20.2… INTERPRET
With w = .7, we have very little smoothing.

With w = .2, we have too much smoothing.
Trend and Seasonal Effects…
A trend can be linear or nonlinear (or, in fact, take any
number of functional forms).
The easiest way of measuring the long-term trend is by

regression analysis, where the independent variable is time.
Linear Trend Quadratic Trend
Seasonal Analysis…
Seasonal variation may occur within a year or within shorter
intervals, such as a month, week, or day.
To measure the seasonal effect, we compute seasonal

indexes, which gauge the degree to which the seasons differ
from one another.
Employment numbers, for example, are seasonally

adjusted to account for summer jobs of students, etc. Was
the change in employment numbers due to seasonality or a
real change in the economy?
Computing Seasonal Indexes…
We can use this procedure to compute seasonal indexes:
u Compute the sample regression line:
v For each time period, compute the ratio:
w For each type of season, compute the average of the ratios

from step v
x Adjust the averages in w so the average of all seasons = 1
Example 20.3… Compute
Calculate the seasonal indexes to account for variations in

Bermuda hotel occupancy rates from Xm20-03:
Year Rate Quarter Time Y-hat
2009 0.561 1 1 0.645
0.702 2 2 0.650 Add the independent variable, time…
0.800 3 3 0.655
2010
0.568
0.575
4
1
4
5
0.660
0.666
Compute the sample regression line:
0.738 2 6 0.671
0.868 3 7 0.676
0.605 4 8 0.681 Regression
2011 0.594 1 9 0.687 Analysis…
0.738 2 10 0.692
0.729 3 11 0.697
2012
0.600
0.622
4
1
12
13
0.702
0.708
Add a new spreadsheet column…
0.708 2 14 0.713
0.806 3 15 0.718
0.632 4 16 0.723
2013 0.665 1 17 0.729
0.835 2 18 0.734
0.873 3 19 0.739
0.670 4 20 0.744
For each time period, compute the ratio: Occ. Rate/Y-hat

Year Rate Quarter Time Y-hat Ratio
2009 0.56 1 1 0.645 0.870
0.7 2 2 0.650 1.080
0.800 3 3 0.655 1.221
0.57 4 4 0.660 0.860
2010 0.58 1 5 0.666 0.864
0.74 2 6 0.671 1.100
0.87 3 7 0.676 1.284
0.61 4 8 0.681 0.888
2011 0.59 1 9 0.687 0.865
0.74 2 10 0.692 1.067
0.73 3 11 0.697 1.046
0.600 4 12 0.702 0.854
2012 0.62 1 13 0.708 0.879
0.71 2 14 0.713 0.993
0.81 3 15 0.718 1.122
0.63 4 16 0.723 0.874
2013 0.67 1 17 0.729 0.913
0.84 2 18 0.734 1.138
0.87 3 19 0.739 1.181
0.670 4 20 0.744 0.900
For each type of season, compute the average of the ratios

from step
2009
2010
2011
2012
2013
Adjust the averages so the average of all seasons = 1
2009
2010
2011
2012
2013
No adjustments are required since:
Average of
all seasons
Alternatively, you can set up the data in this fashion:
[ yt ] [season code]
and use the

Seasonal Indexes tool
from Data Analysis Plus
Example 20.3 INTERPRET
The seasonal indexes tell us that, on average, the occupancy

rates in the first and fourth quarters are below the annual
average, and the occupancy rates in the second and third
quarters are above the annual average.
E.g., we expect the occupancy rate in the first quarter to be

12.2% (100% - 87.8%) below the annual rate, and 7.6%
above for the second quarter, etc.
Time Series and Trend…
Here is the time series data and the regression line together:
Deseasonalizing a Time Series…
One application of seasonal indexes is to remove the
seasonal variation in a time series, by deseasonalizing. The
result is called a seasonally adjusted time series. This allows
us to more easily compare the time series across seasons…
Actual Time Series

Seasonally Adjusted Time Series = Season Index
2009
2009
2009
200
2010
2010
Effects of “Deseasonalization”…
Here we’re comparing the original occupancy rate time
series data with the seasonally adjusted time series data:
Interpretation…
Compared to a horizontal line, we can see that occupancy
rates are rising over time…
Introduction to Forecasting…
There are many different forecasting models available to us.
One way to choose with method or model to use is to select
the technique with the greatest forecast accuracy. Two
measures of this quantity are:
Mean Absolute Deviation (MAD):
and Sum of Squares for Forecast Error (SSE):
(yt = actual value of time series at time t, Ft = forecasted value, n =

number of time periods)
Which to use? SSE? MAD?
MAD averages the absolute differences
between the actual and forecast values.
SSE is the sum of the squared differences.
Which measure to use in judging forecast accuracy depends

on the circumstances:
 If avoiding large errors is important SSE should be used

because it penalizes large deviations more heavily than does
MAD. Otherwise use MAD.
Model Selection…
Here is a useful procedure for model selection:
u Use some of the observations to develop several

competing forecasting models.
v Run the models on the rest of the observations.
w Calculate the accuracy of each model.
x Select the model with the best accuracy measure
Example 20.4…
We have developed three forecasting models; which model
performed best?
2010
2011
2012
2013
E.g. Actuals vs. Forecast #1…
2010
2011
2012
2013
Model 2 is inferior to both models 1 and 3 – drop it.
Using MAD, model 3 is best, but using SSE, model 1 is most

accurate. So? Which one to choose?!
2010
2011
2012
2013
The choice between model 1 and model 3 should be made on

the basis of whether we prefer a model that consistently
produces moderately accurate forecasts (model 1) or one
whose forecasts come quite close to most actual values but
miss badly in a small number of time periods (model 3).
Forecasting Models…
The choice of a forecasting technique depends on the
components identified in the time series. Three techniques
will be discussed…
Exponential smoothing,
Seasonal indexes, and
Autoregressive models.
Forecasting with Exponential Smoothing…
IF the time series
— displays a gradual or no trend, and
— no evidence of seasonal variation,
THEN exponential smoothing can be effective as a
forecasting method.
The forecast for period t+k (k=1, 2, 3,…) is given by:

Ft+k = St
where St is the exponentially smoothed value computed

using techniques discussed earlier.
Forecasting with Exponential Smoothing…
k=1 Ft+1 = St
k=2 Ft+2 = St
k=3 Ft+3 = St
As you can see, we can produce a reasonably accurate

prediction for the next time period (t+1), but the accuracy of
the forecast decreases rapidly more than one time period into
the future (i.e. t+2, t+3, …)
Forecasting with Seasonal Indexes…
IF the time series
— is composed of seasonal variation, and
— has a long-term trend,
THEN we can use seasonal indexes and the regression
equation to forecast.
The forecast for time period t is:

Ft = [b0 + b1t] x SIt
Example 20.5…
Forecast hotel occupancy rates for the next year in Example
20.3…
We know…
the regression line:
and
the Seasonal Indexes:
Put them all together using: Ft = [b0 + b1t] x SIt
Example 20.5…
Ft = [b0 + b1t] x SIt
Continuing into the next year (through the first to fourth

quarter, i.e. time periods 20+1, 20+2, 20+3, and 20+4):
That is, we forecast the quarterly occupancy rates to be:

.658, .812, .890, and .670
Autoregressive Model…
IF the time series
— has no obvious trend or seasonality, but
— we believe that there is a correlation between
consecutive residuals
THEN the autoregressive model may be most effective.
The autoregressive forecasting model is given by:
and is estimated using the regression line:
Example 20.6
The consumer price index (CPI) is a general measure of
inflation and is widely used. Consider the annual percent
increases in CPI collected over 33 year years and forecast
next year’s change in the CPI.
Example 20.6
Year CPI % Change Year CPI % Change
1980 82.4 1997 160.5 2.3%
1981 90.9 10.4% 1998 163.0 1.5%
1982 96.5 6.2% 1999 166.6 2.2%
1983 99.6 3.2% 2000 172.2 3.4%
1984 103.9 4.4% 2001 177.0 2.8%
1985 107.6 3.5% 2002 179.9 1.6%
1986 109.7 1.9% 2003 184.0 2.3%
1987 113.6 3.6% 2004 188.9 2.7%
1988 118.3 4.1% 2005 195.3 3.4%
1989 123.9 4.8% 2006 201.6 3.2%
1990 130.7 5.4% 2007 207.3 2.9%
1991 136.2 4.2% 2008 215.3 3.8%
1992 140.3 3.0% 2009 214.6 -0.3%
1993 144.5 3.0% 2010 218.1 1.6%
1994 148.2 2.6% 2011 224.9 3.1%
1995 152.4 2.8% 2012 229.6 2.1%
1996 156.9 2.9%
Example 20.6 COMPUTE
Remember, we’re trying to correlate the CPI in time period t with the
previous time period, t–1, hence we modify the dataset from the list
set-up in the first Excel snippet to the set-up in the second (which is
how the dataset Xm20-06 is structured).
Now we can run the Regression tool…
Our regression analysis runs…
ŷ t  .0175  .3910 y t 1
Because the last CPI change is 2.1%, our forecast for 2013 is
ŷ 2013  .0175  .4053 y 2012  .0175  .3910 (2.1)  .84
The autoregressive model forecasts a .84% increase in the

CPI for the year 2013.

Time Series Analysis and Forecasting

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Time Series Analysis and Forecasting

Hochgeladen von

Copyright:

Verfügbare Formate

Statistics

for Management & Economics 10e

Time Series Analysis and

As you might expect, this technique has wide application:

A trend is a long term

A cycle is a wavelike pattern

The seasonal component of

Random variation comprises

We can reduce random variation by smoothing the time

To methods to smooth the data are:

Again, calculating manually can be tedious and error prone.

note how the moving average is

The averaging process removes some of the random variation. If we

What happens when we use even numbers of periods to

To place the moving average in an actual time period, we

Two consecutive moving averages are centered by taking

Exponential smoothing addresses these issues…

St = wyt + (1 – w)St-1 (for t ≥ 2)

St-1 = Exponentially smoothed time series at time t–1

We can calculate these values manually…

Excel > Data > Data Analysis > Exponential Smoothing

With w = .7, we have very little smoothing.

The easiest way of measuring the long-term trend is by

Linear Trend Quadratic Trend

To measure the seasonal effect, we compute seasonal

Employment numbers, for example, are seasonally

u Compute the sample regression line:

v For each time period, compute the ratio:

w For each type of season, compute the average of the ratios

x Adjust the averages in w so the average of all seasons = 1

Calculate the seasonal indexes to account for variations in

For each time period, compute the ratio: Occ. Rate/Y-hat

For each type of season, compute the average of the ratios

Adjust the averages so the average of all seasons = 1

No adjustments are required since:

Alternatively, you can set up the data in this fashion:

and use the

The seasonal indexes tell us that, on average, the occupancy

E.g., we expect the occupancy rate in the first quarter to be

Actual Time Series

Mean Absolute Deviation (MAD):

and Sum of Squares for Forecast Error (SSE):

(yt = actual value of time series at time t, Ft = forecasted value, n =

SSE is the sum of the squared differences.

Which measure to use in judging forecast accuracy depends

 If avoiding large errors is important SSE should be used

u Use some of the observations to develop several

v Run the models on the rest of the observations.

w Calculate the accuracy of each model.

x Select the model with the best accuracy measure

E.g. Actuals vs. Forecast #1…

Model 2 is inferior to both models 1 and 3 – drop it.

Using MAD, model 3 is best, but using SSE, model 1 is most

The choice between model 1 and model 3 should be made on

The forecast for period t+k (k=1, 2, 3,…) is given by:

where St is the exponentially smoothed value computed

As you can see, we can produce a reasonably accurate

The forecast for time period t is:

Put them all together using: Ft = [b0 + b1t] x SIt

Continuing into the next year (through the first to fourth

That is, we forecast the quarterly occupancy rates to be:

The autoregressive forecasting model is given by:

and is estimated using the regression line:

Now we can run the Regression tool…