Discriminant Analysis
|
|
Top Previous Next |
|
Discriminant analysis, or canonical variates analysis, is a standard method for testing the significance of previously-defined groups, identifying and describing which variables distinguish between groups, and producing a model to allocate new samples to a group. DA allows the relationship between groups of samples to be displayed graphically. A key goal of a discriminant analysis is to produce a simple function of the variables to classify the objects to their correct group.
DA is closely related to multivariate analysis of variance (MANOVA). It provides information on the relative importance or contribution of each variable to the group structure and produces a method to allocate new observations to a group. To use DA you must allocate samples to groups. For video demonstrations of how to group samples see the Help: Guides: Grouping from Form and Help: Guides: Grouping from Plots.
Some data sets simply will not work with a Discriminant Analysis, if there is low (or zero) variability in the samples/sites, or if there are close correlations between sites or variables. The program will show two error messages; the first shows where in the data set the problem lies:
while the second states that no analysis is possible:
See also: Discriminant Function Coefficients Fisher's Discriminant Functions
|
