Testing Sample Specifications

Click the Select testing sample button on the Data for analysis tab of the Data for analysis step in Data Miner Recipes to display the Testing Sample Specifications dialog box. Use the options in this dialog box to create a hold-out sample that will only be used to validate trained predictive models. The validation sample will never be presented to the predictive models during training.

Specify testing sample using. Use the options in this group box to specify a method for selecting a validation sample (or none at all).

Variable. Select this option button to identify a sampling variable for selecting cases that will be used in forming the validation sample. You will then need to specify a code for the training sample and a code for the validation sample.

Variable name. This field shows the name of the variable that contains the codes for sampling. The validation sample variable is selected during the Data preparation step.

Code for training sample. Use this drop-down list to specify a code for forming the training sample. Data cases pertaining to that specific code will be used for training the predictive models.

Code for testing sample. Use this drop-down list to specify a code for forming the testing sample. Data cases pertaining to that specific code will be used as the validation (or hold-out) sample. They will not be used in training. Instead they will be used after training, to validate the accuracy of the models.

% of cases. Select this option button to specify the percentage of valid cases in the data set that are to be used as the testing sample. You will then need to specify the percentage of cases to use.

Specify %. Enter a number (or use the microscrolls) to specify the percentage of valid cases to use for the testing sample.

none. This is the default setting for the validation sample. Select this option button to ignore forming the validation sample.

OK. Click OK to accept the changes you have made and return to the Data for analysis tab of the Data for analysis step.

Cancel. Click Cancel to the return to the Data for analysis tab of the Data for analysis step without any changes