1 - p(1)
|
Statistica first fits a model with no interactions, then with all two-way interactions, three-way interactions, etc., until a model that fits at the p-value greater than p(1) has been found. You can change this p-value via the 1 - p(1) box. Note that this first part is skipped if an initial User defined starting model is selected.
|
2 - p(2)
|
Next, Statistica eliminates from the initial model in a stepwise manner all effects that do not contribute significantly to the overall fit at the p(2) level of significance. The logic of the automatic selection algorithm is described in the Log-Linear Overview.
|
Starting model
|
The
Starting model box contains two options:
Automatic selection and
User defined.
If you select
Automatic selection, Statistica will perform the computations as described above after clicking the
OK button. During the backwards elimination stage of the automatic model selection procedure, Statistica will drop out each effect, one by one, and test its statistical significance.
If you select User defined, the
Specify Model to be Tested
dialog box is displayed after clicking the
OK button. The model that you then specify will serve as the initial model from which non-significant effects [as specified by p(2)] will be eliminated. This option is also useful if you want to perform statistical significance tests for all effects in a previously determined model. If User defined is selected, the first step of the automatic selection procedure is skipped.
|
OK
|
Click the
OK button to display a summary of the Automatic Selection of Best Model if Automatic selection is selected as the Starting model, or to display the
Specify Model to be Tested dialog box if User defined is selected as the Starting model.
|
Cancel
|
Click the
Cancel button to return to the
Log-Linear Analysis - Model Specification dialog box.
|