Data Analysis with Microsoft Excel: Updated for Office 2007

Chapter 11 Times Series 477

The standard error of the forecast is 72.60194, indicating that the typical

forecasting error in the time series is about 73 units (the MAD value is lower

with a value of about 57.6 units).

The fi nal estimates for the location and trend values are 3488.428 and

14.1268, respectively. The location value represents the monthly sales after

adjusting for seasonal effects. The trend estimate indicates that sales are in-

creasing at a rate of about $14.13 million per month—hardly a large increase

given the magnitude of the monthly sales.

The output also includes a scatterplot comparing the observed, smoothed,

and forecasted production values. Because of the number of points in the

time series, the seasonal curves are close together and diffi cult to interpret.

To make it easier to view the comparison between the observed and fore-

casted values, rescale the x axis to show only the current year and the fore-

casted year’s values.

To rescale the x axis:

1 Click the plot to select it.

2 Click the Axes button on the Axes group of the Layout tab on the

ChartTools ribbon and then click Primary Horizontal Axis and More

Primary Horizontal Axes Options.

3 Click the Fixed option button for the Minimum scale value and

change it to 130. Click the Close button.

The revised plot appears in Figure 11-33.

Figure 11-32
Three-
parameter
exponential
smoothing
values

sales values

forecasted for the

next year

descriptive statistics and

smoothing values

t

w

c