If you know that some independent variables affect another variable of interest (the dependent variable), use multiple linear regression as the forecasting method for that variable. For example, summer temperatures affect electricity usage because, as it gets hotter, more people run their air conditioning. This means that electricity usage (the dependent variable) is dependent on the temperature (an independent variable).
Predictor follows this process to forecast a dependent variable with regression:
Creates an equation that defines the mathematical relationship between the independent variables and a dependent variable. This is the regression equation.
Forecasts each independent variable by running all the selected time-series forecasting methods for each one and using the best method for each.
Calculates the regression equation with the forecasted independent variable values to create the forecast for the dependent variable.
To use multiple linear regression:
On the Predictor wizard Methods panel, select Multiple Linear Regression.
In the Regression Variables dialog, select dependent and independent variables. For instructions, see Selecting Regression Variables.
Select the regression method to use: Standard, Forward stepwise, or Iterative stepwise. For descriptions, see the Glossary in this document and the Oracle Crystal Ball Reference and Examples Guide.
If you selected a stepwise regression, you can select associated settings.
For instructions, see Setting Stepwise Regression Options.
Select or clear the remaining settings:
Include constant in regression equation—Includes the y-intercept constant in the regression equation; if not selected, the regression equation passes through the origin. This setting is selected by default.
Run only regression method for dependent variables—If selected, forecasting methods other than regression are not run on dependent variables. By default, this setting is not selected and all the forecasting methods run on these variables along with linear regression.
Calculate variance inflation factor (VIF) for independent variables—Calculates the Variance Inflation Factor (VIF) of each independent variable included in the regression equation, where VIF is a measure of the strength of multicollinearity (amount of correlation) between the independent variables. Calculating the VIF requires additional time. By default, this setting is not selected.
Note: | For rules concerning minimum number of data points required for multiple linear regression, see Creating Spreadsheets with Historical Data. |