Skip to main content

Table 2 Comparison of effect on population variance between different normalization strategies

From: Normalization of array-CGH data: influence of copy number imbalances

  P-values for data sets
  Data set 1 [23] 2 [25] 3 [20] 4 [8] 5 6 8 [21]
  Nbr of samples 7 28 10 52 8 8 8
  Platform BAC 32 K BAC 32 K BAC 32 K BAC 1 Mb Agilent 244 K Agilent 44 K BAC 32 K
popLowess vs Lowess All populations 1.1e-4 7.0e-12 5.6e-8 3.4e-28 2.5e-05 1.6e-4 7.2e-5
  Population 1 7.8e-3 1.4e-5 9.8e-4 9.9e-32 2.0e-3 2.0e-3 3.9e-3
  Population 2 7.8e-3 1.5e-6 9.8e-4 7.4e-4 2.0e-3 0.09 3.5e-2
  Population 3 0.23 6.3e-3 2.0e-2 2.5e-7 0.25 0.09 0.25
popLowess vs Median All populations < 1e-32 < 1e-32 < 1e-32 < 1e-32 < 1e-32 < 1e-32 < 1e-32
  Population 1 7.8e-3 3.7e-9 9.8e-4 9.9e-32 2.0e-3 2.0e-3 3.9e-3
  Population 2 6.3e-2 1.4e-5 0.17 3.8e-2 0.50 0.09 0.14
  Population 3 0.23 9.0e-5 0.25 0.28 0.25 0.50 0.75
  1. P-values for different populations for data sets are shown. The test corresponds to the null hypothesis that lower standard deviations for popLowess are obtained by chance. Population 1 always relates to the largest identified population (# probes), population 2 to the second largest and population 3 to the smallest.