Table 8.

Comparison of methods in prediction of ER status.

Method	Training Data (two data sets)
	Sotiriou+Miller	expObreast+Sotiriou	expObreast+Miller
TSP	.869 (.950, .788)	.803 (.676, .930)	.727 (.533, .920)
TST	.867 (.950, .783)	.876 (.882, .869)	.769 (.600, .938)
	Training Data (one data set)
Method	Sotiriou	Miller	expObreast
TSP	.883 (.926, .840)	.847 (.789, .904)	.786 (.641, .932)
TST	.841 (.936, .745)	.852 (.800, .904)	.806 (.688, .924)

Top: classification results when training on the data set obtained by integrating two of the three microarray data sets and testing on the third one. Bottom: classification results when training on one of the three microarray data sets and testing on the data set obtained by integrating the other two. The numbers in the parentheses are the sensitivity and specificity levels respectively.