GDA Models Results
GDA Introductory Overview
Click the OK button on the GDA Quick Specs dialog to display the GDA Models Results dialog. This dialog can contain nine tabs, depending on the type of analysis and whether you have selected the More results button (see below): Quick, Functions, Cases, Effects, Profiler, Regression Residuals 1, Regression Residuals 2, Matrix, and Report. Note that this dialog will also be displayed when you click the OK (Run) button on either the GDA Models Wizard Effects in Design dialog, the GDA Models Wizard Extended Options dialog, or the GDA Models Syntax Editor dialog.
The tabs and options available on this dialog are almost identical to those described for STATISTICA General Regression Models (GRM); see the GRM Results dialog (for an overview, see also Methods for Analysis of Variance). The Quick tab contains summary results options, the Functions tab contains results options for discriminant function analysis, the Cases tab contains classification of cases options, the Effects tab contains tests of effects in the model options, the Profiler tab contains the desirability profiler and response optimization options, the Regression Residuals 1 tab and Regression Residuals 2 tab contain regression-like residual statistics and predicted values options, the Matrix tab contains SSCP matrices, estimable functions, etc., and the Report tab contains options to send results to a report.
As described in the Introductory Overview, GDA analyzes discriminant analysis problems by (dummy-) coding the categorical dependent variable into as many new dependent variables as there are classes (groups), so that one dependent variable is created for each class (group); those coded dependent variables contain 1's and 0's to indicate whether or not a case belongs to the respective class. The problem is then analyzed as a multivariate (multiple dependent variable) regression problem [see General Regression Models (GRM)].
The differences between the results reported by GRM and GDA pertain to the different types of analyses which are available in GDA, and not available in GRM, and vice versa; specifically:
- GDA provides options for reviewing the standard results for discriminant analysis, such as the classification functions, canonical factor structure coefficients and means for canonical variates, etc.; those results statistics are also discussed in the context of the Discriminant Analysis module.
- GDA provides options for computing various classification statistics for discriminant analysis, such as Mahalanobis distances for each observation from each group centroid, posterior classification probabilities, and actual classifications. Like the regression residual statistics in GRM, classification statistics can be computed for the analysis sample, as well as the cross-validation and prediction sample; see also the GDA Quick Specs dialog - Advanced tab.
- In addition to the predicted response/desirability profiler options provided in GRM (and GLM as well as the Experimental Design module), GDA provides options for profiling posterior classification probabilities (instead of predicted responses and desirability scores); this unique option allows you to evaluate how different values for the predictor variables affect the predicted classification of cases, and is particularly useful when interpreting the results for complex models that involve categorical and continuous predictors and their interactions.
- More results/Less
- Click the More results/Less button and the GDA Models Results dialog will either be displayed with additional (or fewer) tabs and options.
- Modify
- Click the Modify button to display the previous dialog for the respective analysis (see Methods for Specifying Designs). You will then be able to modify the current analysis.
- Close
- Click the Close button to close the current results dialog and return to the General Discriminant Analysis (GDA) Models Startup Panel.
- Options
- Click the Options button to display the Options menu.
See also GDA Index and Methods for Analysis of Variance.