Discriminant Analysis

[#1]

Build a boundary based statistical model to predict a categorical outcome (classify) as a function of multiple continuous preditor variables.

Discriminant Analysis

1. From an open JMP^® data table, select Analyze > Multivariate Methods > Discriminant.
2. Select one or more continuous variables from Select Columns, and click Y, Covariates (continuous variables have blue triangles).
3. Click on a categorical variable from Select Columns, and click X, Categories (nominal variables have red bars, ordinal variables have green bars).
4. Click OK.  By default, JMP displays the Canonical Plot and Discriminant Scores.

• The Canonical Plot shows the points and multivariate least-squares means on the first two canonical variables that best separate the groups.
• The Biplot Rays on the Canonical Plot indicate
  the directions of the predictors in the canonical space.
• The Discriminant Scores report shows information used to classify each row in the data table.
• The Score Summaries report provides a summary of the misclassifications and tables that tabulates the number and percent of correctly and incorrectly classified cases.

Tips:

• JMP provides Stepwise Variable Selection and three Discriminant Methods (Linear, Quadratic and Regularized).
• Click on the red triangle to select Stepwise Variable Selection, change the discriminant method, show canonical details, specify prior probabilities, save results, customize plots or select other options.
• If a validation column is specified in the model dialog, the Score Summaries table will include counts and misclassification rates for the training, validation (and test) partitions.

Iris.jmp (Help > Sample Data Folder)

Visit Multivariate Methods > Discriminant Analysis in JMP Help to learn more.