Table 2

Assessment of algorithm performance on data simulated according to the heteroscedastic error model (Equation 26)



Power
 Rate of false positives
 RMS bias (×10-2)



(%)
 f
 q
 Naive
 NoSeCoLoR
 Naive
 NoSeCoLoR
 5th percentile
 95th percentile
10
 1.5
 0
 0.312
 0.346
 1.577
 0.890
 0.933
 1.669
10
 1.5
 1
 0.130
 0.342
 0.775
 0.784
 16.536
 17.763
10
 2.5
 0
 0.982
 0.939
 1.482
 0.970
 1.474
 3.447
10
 2.5
 1
 0.683
 0.939
 0.749
 0.855
 15.740
 17.271
20
 1.5
 0
 0.313
 0.345
 1.600
 0.878
 0.930
 2.091
20
 1.5
 1
 0.128
 0.324
 0.784
 0.803
 16.320
 17.722
20
 2.5
 0
 0.983
 0.905
 1.560
 1.367
 3.113
 5.967
20
 2.5
 1
 0.685
 0.909
 0.751
 1.078
 15.299
 16.821

The proportion, , among all genes of those for which the expression level has been changed is either 10% or 20%. The ratio, f, of treated expression level to mean control expression level is varied between 1.5 and 2.5. The bias multiplier q is either zero (no bias) or 1 (bias as measured in the analysis of the real data). The power is the mean number of correct discriminations achieved in the test divided by the number of true changes (59 and 119 for = 10% and = 20%, respectively). The false-positive score is the mean number of incorrect discriminations divided by the expected number at the nominal type-I error rate of 0.01. The expected number of false positives is 5.4 when = 10% and 4.8 when = 20%. The RMS bias is computed from the bias as estimated as described in the text. Reported here are the 5th and 95th percentiles over the simulated datasets.

Kepler et al. Genome Biology 2002 3:research0037.1   doi:10.1186/gb-2002-3-7-research0037