Discriminant Node Expert Options
If you have detailed knowledge of discriminant analysis, expert options allow you to fine-tune the training process. To access expert options, set Mode to Expert on the Expert tab.
Prior Probabilities. This option determines whether the classification coefficients are adjusted for a priori knowledge of group membership.
- All groups equal. Equal prior probabilities are assumed for all groups; this has no effect on the coefficients.
- Compute from group sizes. The observed group sizes in your sample determine the prior probabilities of group membership. For example, if 50% of the observations included in the analysis fall into the first group, 25% in the second, and 25% in the third, the classification coefficients are adjusted to increase the likelihood of membership in the first group relative to the other two.
Use Covariance Matrix. You can choose to classify cases using a within-groups covariance matrix or a separate-groups covariance matrix.
- Within-groups. The pooled within-groups covariance matrix is used to classify cases.
- Separate-groups. Separate-groups covariance matrices are used for classification. Because classification is based on the discriminant functions (not based on the original variables), this option is not always equivalent to quadratic discrimination.
Output. These options allow you to request additional statistics that will be displayed in the advanced output of the model nugget built by the node. See the topic Discriminant Node Output Options for more information.
Stepping. These options allow you to control the criteria for adding and removing fields with the Stepwise estimation method. (The button is disabled if the Enter method is selected.) See the topic Discriminant Node Stepping Options for more information.