False positive rates across concentration. False positive rates across concentration using a p < 0.05 threshold. Genes are partitioned into 20 equal-sized bins ordered on expression intensity; for each bin the FPR is calculated by dividing the number of significant genes in the bin by the total number of genes in the bin. (A) HGU133A Latin Square data with "spiked-in" genes removed. The dashed line is the expected 0.05 FPR rate. (B) Simulated local variance model null hypothesis data with high variance (n = 3). Differential expression tests: FC: Fold Change, limma: limma moderated t-statistic, BRB: BRB t-statistic, CyberT: CyberT t-statistic, LPE: LPE z-statistic, t-stat: t-statistic. The limma and BRB t-statistic produce comparable results and consequently limma's orange line often conceals BRB's green line.