Figure 2
From: Empirical Bayes models for multiple probe type microarrays at the probe level
False positive rate against mean intensity. False positive rate (α) calculated on re-sampled data and plotted against mean intensity. 100 data sets of size 6 were sampled from the complete data set B (see Section Data sets) of 18 replicated arrays and then analyzed using the Affymetrix MAS5 algorithm followed by a two group analysis of 3+3 arrays using the moderated t-test in the R-package LIMMA [3], on logged MAS5 indexes and indexes transformed using the variance stabilizing transformation in the R-package vsn [21], and the proposed method LMW using logged MAS5 indexes. false positive rate were obtained by averaging over the sampled data sets using loess-curves fitted to mean intensity and indicator of significance (1 if the probe-set is among the 5% probe-sets with highest absolute statistic, 0 otherwise). The mean intensities of each data set are shifted to the range [0,15].