Detection of Multicollinearities

Command:

Math -> Multiple Regression -> Detection of Multicollinearities...

The command Math/Multiple Regression/Detection of Multicollinearities... (toolbar button ) supports the detection of multicollinearities by means of the variance inflation factor (VIF). The user has to select the variables to be included by ticking off the corresponding check boxes. In general one starts with the selection of all variables, and proceeds by successively deselecting variables showing a high VIF. Ideally, the VIF values should be below 10. The selected variables can be transfered to the MLR window by clicking the button.

Hint 1:

For datasets with less than 25 variables the calculation of the VIF values is performed automatically on any change of the variable selection. If a dataset contains many more variables the immediate calculation may take considerable time which proves to be a hinderance. In this case the button "start the calculation" is displayed and the user has to start the VIF calculation explicitely by clicking this button.

Hint 2:

An infinite VIF value indicates that the corresponding variable may be expressed exactly by a linear combination of other variables (which show an infinite VIF, as well).

Hint 3:

If the number of selected variables is greater or equal than the number of observations the VIFs cannot be calculated. In this case no VIF values are displayed and a remark is added to the protocol.