Distribution of actual test statistics vs. randomly permuted background. Distribution of ANOVA F-statistics from the model shown in Figure 1 were calculated for the observed (FStat Observed) and permuted datasets. The maximum, 95th percentile, and 90th percentile F-Statistics in the permuted dataset provide an index of the distribution of test results expected for a random sample. The values of test statistics seen at the 95% level in a randomly permuted dataset are greater than those of the observed dataset. Genes were included if at least one tissue showed expression above LN(100) (7853 genes).
Mutter et al. BMC Genomics 2004 5:88 doi:10.1186/1471-2164-5-88