For citing, please use (manuscript submitted).

Name: Wrage et al. (30 arrays)Group A: BM+ (13)FR40_N9_S01_output, FR40_S2_S01_output, MK_2748_S01_output, MK_A16_S01_output, MK_D5_S01_output, MK_E6_S01_output, MK_F1_S01_output, MK_H17.hell_S01_output, MK_H18_S01_output, MK_S10_S01_output, MK_S22_S01_output, MK_S28_S01_output, PE_R21.Omni31.slide.33_S01_outputGroup B: BM- (15)X4105_output, FR40_3726_S01_output, FR40_F7_S01_output, MAF_Omni43slide48_3325_S01_output, MAF_Omni43slide49_3201_S01_output, MAF_Omni43slide51_4462_S01_output, MK_2603_S01_output, MK_3121_S01_output, MK_3215_S01_output, MK_3245_S01_output, MK_4298_S01_output, MK_4401_S01_output, MK_B4_S01_output, MK_H14_S01_output, MK_N20_S01_outputUnused: (2)MAF_Tumor4105.Omni31.slide.34_S01_output, MK_G8_S01_output

Average power plotted as a function of sample size.

Copy number profiles produced with CGHcall. The mean probability of losses is shown in red, and the values can be red from the Y axis. Mean probability of gains is in green, and the values are 1 - the value from the Y axis. Output from CGHregions. Chromosomes are plotted individually, and each bump represents a breakpoint between regions. The loss/gain frequencies are shown in red/green.

To help interpreting the these plots, comparison can be made to the evaluation data sets. The two estimators of G should not be in too much of a disagreement compared to each other. If the difference is severe, the quality of parameter estimation is questionable, and so is the reliability of the power calculations. The density of the p values should increase for small values and the function should be convex. Gamma is the proportion of non-differentially behaving regions. The plot shows the distance between the two estimators of G, and the value of gamma is chosen so that this difference is minimized. The function should have a minimum somewhere in the middle. A minimum very close to 0 is a sign of problems in parameter estimation. Effect size is the difference between the groups. The function can have one ore more peaks, depending on the particular data set. Skewness of the RWLRs is plotted and that of a normal distribution is superimposed in red. RWLRs are assumed to be approximately normally distributed, so a large deviation might negatively affect the performance of the method. Kurtosis of the RWLRs is plotted and that of a normal distribution is superimposed in red. RWLRs are assumed to be approximately normally distributed, so a large deviation might negatively affect the performance of the method.