By default, the Predictor Results window contains a chart of historical and forecasted values in the upper left.
To control the chart view, use these settings:
Periods to forecast— Determines the number of forecasted values that are displayed in the chart
Confidence interval—Indicates which confidence interval to calculate and plot
Series—Selects the data series to display in the chart
Method—Selects the method to use for calculating forecasted values
View menu—View, Table changes the chart display to a table; View, Chart changes it back; and View, Show Statistics hides and displays the statistics tables to enlarge the chart
Note:
If Include events is selected in the Data Attributes panel of the Predictor wizard, and at least one event is defined, Table View includes an Event column with the name and number of each event defined for the selected series.
Preferences menu—Preferences, Chart displays the Chart Preferences dialog (see Customizing Charts, following); Preferences, Show All Error Measures hides and shows error measures that are not selected in the Options panel of the Predictor wizard; Preferences, Highlight Seasonality graphically emphasizes seasonal data cycles if present; Preferences, Highlight Screened Data emphasizes filled-in or adjusted-outlier data if these are present and you selected at least one of the Data Screeningsettings in the Data Attributes panel; and Preferences, Highlight Events emphasizes data defined as events if you have defined at least one event and selected Include events in the Data Attributes panel.
As shown in the chart legend, the green line represents the historical data, the blue lines represent fitted and forecasted values, and the red dotted lines above and below the forecasted values represent the upper and lower confidence interval. A gap between the historical and forecasted values delineates the past and future values.
Of the eight classic time-series forecasting methods, four result in straight lines: single moving average, single exponential smoothing, double moving average, and double exponential smoothing. Only the seasonal methods and multiple linear regression result in curves that approximate repeated data patterns.