17.7.4.1 The Discriminant Analysis Dialog Box

In the Discriminant Analysis dialog, you can perform Discriminant Analysis or Canonical Discriminant Analysis. To perform Canonical Discriminant Analysis, click the button Button Select Dialog Theme.png to the right of Dialog Theme and select the Canonical Discriminant Analysis theme from the menu.

Recalculate

Specify the way to recalculate and update the result if there is any change in the input data or settings.

None The output will not be connected to the source data, and any change will not result in an update of the result. You can't change settings to recalculate the result.
Auto The result automatically updates when source data changes. You can also change settings to recalculate the result.
Manual The result will not automatically update when source data changes. You must manually activate the update by clicking the Recalculate button Button Recalculate Manual.png in the Standard toolbar. You can also change settings to recalculate the result.

Input

Select data for Discriminant Analysis.

Group for Training Data Select data from a column to specify group for training data. Note that the grouping column will be set as categorical if Text column.
Training Data Select data to specify training data.

Note that the number of rows of Group for Training Data and Training Data should be the same, otherwise only the rows in Group for Training Data corresponding to Training Data are included in analysis.

Predict Membership for Test Data Determine whether to predict membership for test data. If checked, Test Data control is shown.
Test Data Select data to specify test data.

Note that test data should contain the same number of variables as training data.

Settings

Specify the settings in Discriminant Analysis.

Prior Probabilities Select the type of prior probabilities for each group. Origin supports two types:
  • Equal
Prior probabilities are equal for all groups.
  • Proportional to group size
Prior probability for a group is proportional to the number of observations in the group.
Discriminant Function Select the method to classify. Origin provides two methods:
  • Linear
Using Linear Discriminant Function. The pooled within-group covariance matrix is used to calculate Mahalanobis distance.
  • Quadratic
Using Quadratic Discriminant Function. Within-group covariance matrices are used to calculate Mahalanobis distance.

For more details, see the algorithm of discriminant functions.

Canonical Discriminant Analysis Specify whether to perform Canonical Discriminant Analysis. If not selected, Canonical Discriminant Analysis branch in Quantities group and Canonical Score Plot check box in Plots group will be disabled.
Cross Validation Specify whether to classify training data using Cross Validation method.

Statistics

Specify whether to perform statistics analysis on training data (e.g., Descriptive Statistics, Univariate ANOVA).

Descriptive Statistics Specify whether to perform Descriptive Statistics on training data, including means, standard deviations for each variable in each group, and total.
Descriptive Matrices Specify whether to calculate covariance matrix, correlation matrix and group distance(squared Mahalanobis) matrices of training data.
Univariate ANOVA Specify whether to perform Univariate ANOVA on training data to test the difference in group means for each variable.
Equality Test of Covariance Matrices Specify whether to perform Equality Test of Covariance Matrices on training data to test the equality of within-group covariance matrices.
Pooled Within-group Covariance/Correlation Matrix Specify whether to output pooled within-group covariance matrix and correlation matrix for training data.
Within-group Covariance Matrices Specify whether to output within-group covariance matrices for training data.

Quantities

Specify quantities to calculate in Discriminant Analysis.

Discriminant Function Coefficients Specify whether to calculate discriminant function coefficients including constant and linear coefficients. Enabled only when Linear is chosen in the Discriminant Function radio box.
Canonical Discriminant Analysis This branch determines which quantities to calculate in Canonical Discriminant Analysis. It includes the following check boxes.
  • Canonical Structure Matrix
Specify whether to calculate canonical structure matrix in Canonical Discriminant Analysis.
  • Canonical Coefficients
Specify whether to calculate canonical coefficients in Canonical Discriminant Analysis, including unstandardized canonical coefficients and standardized canonical coefficients.
  • Canonical Scores
Specify whether to calculate canonical scores and canonical group means in Canonical Discriminant Analysis. Note that when Canonical Score Plot is seelected in Plots group, it will be checked and disabled automatically.

Note that Eigenvalues, Wilks' Lamba Test will be included in the result of Canonical Discriminant Analysis by default.

For details, see the introduction of Canonical Discriminant Analysis.

Classification Results This branch determines which quantities are included in the classification result of training data, test data, and cross validation result of training data. It includes the following check boxes.
  • Posterior Probabilities
Specify whether posterior probabilities are included in the classification result for observations of training data and test data in different groups. Note that when Classification Fit Plot is chosen in Plots group, it will be checked and disabled automatically.
  • Squared Mahalanobis Distance
Specify whether squared Mahalanobis distance is included in the classification result for observations of training data and test data in different groups.
  • Atypicality Index
Specify whether atypicality index is included in the classification result for observations of training data and test data in different groups.

Note that predicted membership for each observation is listed in the classification result by default.

For details, see the introduction of classification result.

Classification Summary Specify whether to summarize the classification result, including observation count in each predicted group, error rate for training data and cross validation of training data. Note that if Classification Summary Plot is selected in Plots group, it will be chosen and disabled automatically.

Plots

Specify whether to show plots in Discriminant Analysis report.

Classification Summary Plot Specify whether to show Classification Summary Plot in the report, which shows the source of predicted group members.
Classification Fit Plot Specify whether to show Classification Fit Plot in the report, which shows the posterior probabilities of observations for the predicted group.
Canonical Score Plot Specify whether to show Canonical Score Plot in the report, which shows scores of observations in the first two canonical variables.

Output

Specify the destination of output results.

Discriminant Analysis Report Specify the sheet for the discriminant analysis report. The default value is a new sheet in the workbook of input data.
Classification Result for Training Data Specify the sheet for the classification result of training data. The default value is a new sheet in the workbook of input data. Note that the sheet will not be created for Canonical Discriminant Analysis if it is set to <optional>.
Classification Result for Test Data Specify the sheet for the classification result of test data. The default value is a new sheet in the workbook of input data. Note that it will be disabled if Predict Membership for Test Data is not selected in Input Data group.
Canonical Scores Specify the sheet for canonical scores. The default value is a new sheet in the workbook of input data. Note that it will be disabled if neither Canonical Scores in the Canonical Discriminant branch of Quantities group nor Canonical Score Plot in Plots group is selected.