ChipStat behavior using simulated biological/experimental + hybridization noise model. The behavior of the ChipStat algorithm was evaluated (pps = 0.05, 16 probe pairs per probe set) using a Monte Carlo model in which the ratio of biological + experimental noise (E(b) + E(p)) to hybridization noise (E(h)) is constant (see text for further details). Results are shown for E(h) = 0 (Exp noise only; blue), E(h) = E(b) + E(p) (Hyb noise = Exp noise; red), E(h) = 2 × (E(b) + E(p)) (Hyb noise = 2 × Exp noise; green), and E(b) + E(p) = 0 (Hyb noise only; yellow). The total number of probe sets simulated (11,820) was chosen to match the number of probe sets containing 16 probe pairs per probe set on the Affymetrix MG_U74Av2 array. The number of probe pairs increasing by chance is shown on the x axis, and the fraction of total probe sets simulated is shown on the y axis. This simulation was repeated 100×, and the average of these results is shown. (a) Probability of the indicated number of probe pairs increasing. (b) Cumulative P value (equal to or greater than the indicated number of probe pairs changing).
Master et al. Genome Biology 2005 6:R20 doi:10.1186/gb-2005-6-2-r20