This is an archived page, please click here to view the latest updates.

OriginLab Corporation - Scientific Graphing and Data Analysis Software - 2D graphs, 3D graphs, Contour Plots, Statistical Charts, Data Exploration, Statistics, Curve Fitting, Signal Processing, and Peak Analysis

Discriminant Analysis


Contents


Discriminant analysis is used to distinguish distinct sets of observations and allocate new observations to previously defined groups. This method is commonly used in biological species classification, in medical classification of tumors, in facial recognition technologies, and in the credit card and insurance industries for determining risk.

Goals

There are two main goals for discriminant analysis:

Assumptions

The discriminant model has the following assumptions:

Processing Procedure

Preparing Analysis Data

Verifying Assumptions

The normality test, Equality Test of Covariance Matrices, and pooled within-groups correlation matrix can be used to verify the assumptions. Please see Assumptions for more information.

Selecting Discriminant Methods

Note: LDA is Linear for Discriminant Function option and QDA is Quadratic for Discriminant Function option

Interpreting and Verifying the Results

See the Interpreting Results page for information on editing discriminant functions, judging whether the discriminant functions are good or not, and classifying observations.

To verify the results, we can judge from the result of test data and cross validation of training data. However, please note that both methods are sensitive to sample size. If the sample size is small, the result may not be reliable.

Note: Cross-validation is also called leave-one-out cross validation. If we have N observations, discriminant analysis will run N times. Each time, the analysis is trained on all data except one point, and a prediction is made on that point.

Performing Discriminant Analysis


Topics covered in this section: