Additional file 4.Estimation of False Discovery Rate (FDR). Monte-Carlo simulation for the diurnal signature was carried out by doing 100 random permutations on the gene expression data and performing a 3-way ANOVA. Panel A illustrates the relationship between the number of probes significant at given p-value level vs. the alpha level as observed on the original (non-randomized) data (red line) and the mean of 100 randomizations (blue line). Panel B depicts the estimated q-value vs. the p-value cut-off. Panel C displays the number of significant probes vs. the estimated q-value for a given p-value significance threshold. Panel D shows the relationship between the estimated mean number of false positives vs. the number of significant probes for a given p-value cut-off. Data are shown for the range of p-values up to 0.05 for all 4 panels. Format: TIFF Size: 427KB Download file Loboda et al. BMC Medical Genomics 2009 2:7 doi:10.1186/1755-8794-2-7 |