Faq how do I interpret the misclassification results reported by crossval

From Eigenvector Research Documentation Wiki
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

Issue:

How do I interpret the Misclassification results reported by crossval?

Possible Solutions:

The Crossval function (used for cross-validation) allows "discriminant analysis" in which a y-vector or matrix is supplied which indicates which samples are in one or more classes. This y-block is usually a logical (boolean) array where each column represents one classes membership. A value of 1 in a column indicates that the given sample is a member of that column's class. A value of 0 indicates that sample is not a member of the class.

The report at the end of crossval provides a tabular description of the results each column. In these tables, the numbers represent "misclassification rates". These are fractional errors of classification where 0 indicates that no samples in the given group were mis-classified and 1 indicates that all samples in the given group were mis-classified.

Specifically, the groups in each table are usually labeled as "class 0" and "class 1" (see below for an example). Class 0 represents the group of samples which were labeled 0 ("not-in-class") for the given column. Class 1 represents the group of samples which were labeled 1 ("in-class"). As such, the misclassification results for Class 0 can be interpreted as false-postive rates and the misclassification results for Class 1 can be interpreted as false-negative rates.

In the example below, the false positive rate for 3 latent variables (components) is 0.076 = 7.6% false positive rate. The false negative rate at 3 latent variables is 0.000, or 0 false negatives (perfect classification).

Fractional Misclassification (Y-column 1)
Comp # Class #
0 1
1 0.530 0.0556
2 0.258 0.111
3 0.076 0.000

Note that these false positive and false negative rates can be easily used to calculate sensitivity and specificity using the relationship:

specificity = 1 - (false positive rate)
sensitivity = 1 - (false negative rate)


Still having problems? Please contact our helpdesk at helpdesk@eigenvector.com