Chapter 11 Times Series 455
Choosing a Value for w
As you saw in the Exponential Smoothing workbook, you have to choose the
value of w with care. When choosing a value, keep several factors in mind.
Generally, you want the standard error of the forecasts to be low, but this is
not the only consideration. The value of w that gives the lowest standard
error might be very high (such as 0.9) so that the exponential smoothing does
not result in very smooth forecasts. If your goal is to simplify the appearance
of the data or to spot general trends, you would not want to use such a high
value for w, even if it produced forecasts with a low standard error. Analysts
generally favor values for w ranging from 0.01 to 0.3. Choosing appropriate
parameter values for exponential smoothing is often based on intuition and
experience. Nevertheless, exponential smoothing has proved valuable in
forecasting time series data.
The ability to perform exponential smoothing on time series data has
been provided for you with StatPlus. Let’s smooth the temperature data,
using a w value of 0.18.
To create exponentially smoothed forecasts of the mean annual
temperature:
1 Return to the Global Temperature Analysis workbook and go to the
Temperature worksheet.
2 Click Time Series from the StatPlus menu and then click Exponen-
tial Smoothing.
3 Click the Data Values button and select Fahrenheit from the list of
range names. Click OK.
4 Type 0.18 in the Weight box under General Options.
5 Click the Output button and direct the output to a new worksheet
named Smoothed Temperature. Click OK.
The completed dialog box appears in Figure 11-15.
