Definition
Discriminant analysis finds combinations of variables that best separate groups known in advance — healthy from ill, good parts from rejects, three varieties by dimensions — and lets you classify new observations. The output is discriminant functions, a classification matrix with accuracy and a plot of canonical scores.
Multivariate methods
The method assumes multivariate normality and equal group covariance matrices; when violated, choose the quadratic variant or logistic regression, which is more robust to assumptions. Stepwise selection picks the variables that really discriminate.
Judge classification accuracy on test data or by cross-validation (leave-one-out), not on the data the model was estimated from — otherwise you overrate it. Canonical analysis shows how many dimensions are needed to separate the groups.
In Statistica
Statistics → Multivariate Exploratory Techniques → Discriminant Analysis offers stepwise variable selection, classification functions, prior probabilities, a classification matrix and canonical analysis with a score plot; General Discriminant Analysis Models (GDA) add categorical predictors and interactions.
Related terms
- Cluster analysisCluster analysis divides observations into groups (clusters) so that units within a cluster are as similar as possible…
- Logistic regressionLogistic regression models the probability of a binary outcome — a patient responds to treatment or not, a customer…
- Decision treesA decision tree splits data by a sequence of simple rules ("age > 45 and income < 30 thousand") into ever more…
- ROC curve and AUCThe ROC curve shows how a classifier's sensitivity (share of correctly caught positives) and false-positive rate change…
Knowledgebase guides
FAQ
- Discriminant analysis or logistic regression?
- With normality satisfied and small samples discriminant analysis is more efficient; otherwise logistic regression is safer and easier to interpret.
- How do I validate classification accuracy?
- By cross-validation or a holdout test sample; the classification matrix on training data is optimistic.
Try it on your own data
Statistica free for 30 days
Full version, no credit card. Or get a pricing estimate in a minute.
Multivariate methods
Updated: September 2026.