1. Home
  2. Linear Discriminant Analysis Options

Linear Discriminant Analysis Options

Use this to select the output to be generated by a linear discriminant analysis.

Display

Specifies which items of output are to be displayed in the Output window.

Counts of units in each group Numbers of units in each group with a complete set of observations
Latent roots The canonical variate loadings, the latent roots and the trace
Chi-square tests Chi-square tests for significance of roots
Correlations between input variables The within-group correlation matrix of the input variates
Correlations between input and canonical variates The within-group correlations between the input and canonical variates
Canonical variate means Canonical variate scores for the group means
Canonical correlation coefficients Canonical correlation coefficients
Adjustments The adjustments required to the canonical variate scores to centre the results about the origin
Distance between group means The inter-group distances
Unit scores Canonical variate scores for the units
Unit to group mean distances Mahalanobis squared distances between the units and the group means
Group allocations The initial grouping and the allocation of units to groups
Counts of allocations Counts of allocations
Validation error rates Error rates from the validation methods specified in the Validation options

Number of dimensions

Specifies how many latent roots and vectors are printed (and saved); the residuals are formed from the remaining dimensions.

Reallocate units in training set to groups

Controls whether the units in the training set are to be reallocated to groups. If this is not selected then their group values, either displayed or saved, will be missing.

Validation options

These options control how the Validation error rates in Display are calculated. The Validation error rates option must be selected to set these options.

Method

Cross-validation Uses the cross-validation error rate, where one group of the data is left out at a time. The number of groups is specified below, and units are randomly allocated to groups.
Bootstrap Uses the bootstrap error rate, where the observations are resampled and the omitted units are predicted.
Jackknife Uses the cross-validation error rate, where one unit is left out at a time and predicted from the other units.

Number of cross-validation groups

For the cross-validation, this gives the number of groups to which the data will be allocated. Each group is then left out of the analysis, and predicted from the remaining groups. On each simulation the units are randomly assigned to groups.

Number of simulations

For the cross-validation or bootstrap validation error rates, this gives the number of times the resampling will be performed to estimate the error rate. Increasing this will slow the analysis down.

Seed

This gives a seed to initialize the random number generation used for bootstrapping and cross-validation. Using zero initializes this from the computer’s clock, but specifying an nonzero value gives a repeatable analysis.

Graphics

Specifies which graphical outputs are to be produced by the analysis.

Discrimination plot Draws a discrimination plot given by the X Root and Y Root boxes. The maximum dimension that can be displayed is that given above. You can specify to display the means, mean labels, unit scores, group polygons enclosing units and 95% confidence circles around the group means. A 1 dimensional plot can be produced by specifying the X root and setting the Y root to

See also

Updated on April 22, 2025

Was this article helpful?