Table 2.

Operating characteristics for Latin Square dataset

	First 64 ranked probesets
	FP (%)	TN (%)	TP (%)	FN (%)
Combined	12 (0.1)	10545 (99.9)	52 (81.3)	12 (18.7)
MAS5	23 (0.2)	10534 (99.8)	41 (64.1)	23 (35.9)
RMA	14 (0.1)	10543 (99.9)	50 (78.1)	14 (21.9)
dChip	31 (0.4)	10526 (99.6)	33 (51.6)	31 (48.4)

The table presents the operating characteristics of the combined method and of each pre-processing method on the first 64 probesets ranked accordingly to their tail posterior probability. The combined strategy is more able to recognize true positives and true negatives than each single method. Note that FP = FN since the size of the list of differentially expressed probesets is equal to the number of true positives.