Data Analysis with Microsoft Excel: Updated for Office 2007

460 Statistical Methods

The values chosen for w and t are important in determining what the

forecasted value for the time series will be. If we assume that the data will

continue to behave as it did for earlier values, smaller values for w and t

might be used, because those would result in an estimate that has a longer

“memory” of previous values. Let’s see what kind of difference this would

make by reducing the value of t from 0.15 to 0.05.

To reduce the value of t:

1 Repeatedly click the down spin arrow next to the Trend constant

until the value of t equals 0.05. See Figure 11-18.

Figure 11-18

Decreasing

the value of

t to 0.05

a positive trend

is forecast

observations further

back in time are

heavily weighted in

forecasting the trend

With this value of t, the forecasted increase in the time series data changes

to 0.030 points per unit, reflecting the assumption that there will be an

increase in the data similar to what was observed earlier in the time series.

Note that the weights for the trend factor, as shown by the area curve, indi-

cate that older observations are well represented in the forecast. Now let’s

see what would happen if we increased the value of t, focusing more on

short-term trends.

To increase the value of t:

1 Repeatedly click the up spin arrow next to the Trend constant until

the value of t equals 0.40. See Figure 11-19.