Supporting information for Tudor et al. (2002) Proc. Natl. Acad. Sci. USA, 10.1073/pnas.242566899
Fig. 4.
The power of an analysis method depends on the sample size used. Sample sizes between 4 (i.e., 2 control and 2 experimental samples) and 18 samples were simulated and the mean power and number of false positives returned over at least 50 simulations is reported vs. the absolute fold change. Where legibility permits, the individual traces are labeled with sample sizes, otherwise the results varied roughly monotonically with sample size. Presented are the "naïve" (with respect to multiple testing) t test and VERA/SAM error model (7), and the results of correcting the P values from these analyses for multiple testing with the Benjamini-Yekutielli algorithm (6). The simulations presented were performed by using the P135+ Mecp22/y; CamK-Cre93+/o vs. Mecp2+/y hippocampus (n = 18, 4095 genes) experiment; simulations conducted using other datasets were comparable.