Outliers can be established either via standard deviation criteria or by user definition. This topic describes how use standard deviation criteria; see Perform Regression with Outliers Set by User-defined Criteria for the other procedure.

To create outliers using standard deviation criteria and perform a regression analysis:

1. Display the Performance Models window.
2. Right-click the node that represents the data set on which linear regression will be performed and select Analyze Model. The system displays the Analyze Model window with a condition versus age plot of all data points in the data set.
3. Click the Models tab.

4. In the right pane of the Models tab, locate the record showing 0-Selected Model.

   Note: The system only performs regression on the model identified with a Model Run type of 0-Selected Model. If the model so identified is not the one on which you wish to perform regression analysis, right-click the record that shows the desired model and select Move for Analysis. This changes the Model Run type of this model to 0-Selected Model, which allows it to be analyzed.

5. The default setting is for data points beyond two standard deviations to be identified as outliers. If this is okay, go to the next step. If you wish to set a different SE, right-click the record identified in step 4 and select Set Standard Error. The system displays the Set SE dialog box. Change the default value of two to the desired number of standard deviations and select OK to close the dialog box.
6. Right-click the row identified in step 4 and select Define Outliers. The system identifies outliers, and then displays the amount of outliers found in the NUM OUTL column. (Note: The value in the NUM SAMPLE column is also decremented by the amount of outliers found.)

7. Right-click the row identified in step 4 and select Calculate Excluding Outliers. The system performs the regression analysis on the data set (minus the outliers) using the current model. Click the Graph tab to view the results of this analysis.

   Note: Data points are colored gray and the outliers are colored black. Also, outliers are only plotted if the model selected in the Models tab is the one that has the 0-Selected value in the Model Run column.