Eigenvalues - DA
|Top Previous Next|
The Eigenvalues tab screen displays the eigenvalues for the discriminant functions and the correlation between the discriminant scores and the group membership variable. See Printing and exporting text to save or print this table.
Eigenvalues measure the amount of variance in the grouping variable explained by the predictors in the discriminant function. There is one eigenvalue for each discriminant function, so if you only have two groups there will only be 1 eigenvalue, which will account for 100% of the explained variation. The discriminant functions are arranged in terms of their discriminatory power so that the first has the largest eigenvalue. The ratio of the eigenvalues for each discriminant function measures the relative importance of each function in discriminating between the groups. In the example above, for example, the largest function (eigenvalue 32.1919, proportion 0.9912) has by far the greatest power to discriminate between the groups.
Proportion is the fraction of the total sum of the eigenvalues represented by any one eigenvalue.
Canonical Correlation is a measure of the association between the groups and those defined by the discriminant function. If there are two groups, then the canonical correlation is the Pearson correlation between discriminant scores and the group membership variable (eg group 1 = 1, group 2 = 2 etc.). It measures the usefulness of the function in discriminating between the groups. A value of zero indicates no relationship and no discriminatory ability. A value of 1 indicates that all the variability in the discriminant scores for objects is generated by a single discriminant function.
Chi-squared is the test statistic for the significance of the observed canonical correlation.
D.F. is the degrees of freedom for the Chi-squared test statistic.
Prob. is the probability that a correlation of the size observed could be generated by random chance.