Suggests golubEsets (> = 1.4.2), vsn (> = 1.7.2)

Description In a typical microarray setting with gene expression data observed under two condi-tions, the local false discovery rate describes the probability that a gene is not differentially ex-pressed between the two conditions given its corrresponding observed score or p-value level. The resulting curve of p-values versus local false discovery rate offers an in-sight into the twilight zone between clear differential and clear non-differential gene expres-sion. Package ’twilight ’ contains two main functions: Function twilight.pval performs a two-condition test on differences in means for a given input matrix or expression set and com-putes permutation based p-values. Function twilight performs a stochastic downhill search to es-timate local false discovery rates and effect size distributions. The package further pro-vides means to filter for permutations that describe the null distribution correctly. Using fil-tered permutations, the influence of hidden confounders could be diminished