Statistics and Data Analysis for Microarrays Using R and Bioconductor, 2nd Edition

956 Statistics and Data Analysis for Microarrays Using R and Bioconductor

discuss linear and quadratic discriminants while from the nonparametric

one, we will describe the k-nearest ne ighbor technique.

The seco nd approach is to use data to estimate the class boundaries di-

rectly, without explicitly calculating the probability dens ity functions. Ex-

amples of algorithms in this category include decision trees and support

vector machines.

29.3.2 Error estimation and validation

Suppose the classifier C(x) was trained to classify input vectors x into two

distinct classes, 1 and 2. The results of the classification of a collection of

input objects x

i

, i = 1, . . . , n ca n be summarized in a confusion matrix

that co ntrasts the predicted class labe ls