Skip to main content

Table 2 Area under ROC curves up to 100 false positives, MAS5, PPLR, BGX, and Logit-T

From: Empirical Bayes models for multiple probe type microarrays at the probe level

Method Pre-processing Affymetrix U95 Affymetrix 133A Golden Spike Gene Logic Tonsil Gene Logic AML
LMW MAS5 89(1) 87(1) 60(1) 79(1) 70(2)
IBMT MAS5 87(2) 87(2) 59(2) 77(3) 69(3)
LPE MAS5 84(3) 84(3) 57(3) 78(2) 79(1)
WAME MAS5 71(6) 81(5) 34(5) 69(4) 54(8)
SAM MAS5 74(4) 81(4) 11(8) 67(5) 54(9)
LIMMA MAS5 71(7) 81(6) 31(6) 67(6) 54(6)
Shrink-t MAS5 71(8) 80(7) 23(7) 67(7) 54(7)
t-test MAS5 73(5) 76(8) 39(4) 60(9) 47(10)
Efron-t MAS5 65(9) 72(9) 3(9) 66(8) 57(4)
FC MAS5 56(10) 61(10) 0(10) 58(10) 55(5)
BGX   - - 58 - 75*
logit-T   94 92 - 80 79
PPLR multi-mgMOS 88 90 57 71 69
# of genes   12626 22029 14010 12626 12626
# of spikes   16 42 1331 11 11
# of groups   20 14 2 12 10
  1. Area under ROC curves up to 100 false positives rounded to nearest integer value with an optimum of 100. Numbers within parenthesis are within data set ranks for the methods compared, and methods are ordered with respect to mean rank across the five data sets (within MAS5 only). *Results in italic are from the subset of 1011 probe sets of the Gene Logic AML data set.