Correlation of observed with actual fold changes for a representative expression summary dataset (Additional data file 2, using dataset 9e.b). (a) The fold change for each probe set with spiked-in target RNA is depicted as a cross. Empty probe sets are not shown. For each actual fold-change level (on the x axis), a boxplot shows the distribution of the corresponding observed fold changes. A linear fit of the data is shown in cyan. Fit parameters: R2 = 0.508; slope = 0.505; y-intercept = -0.061. (b-d) Increasingly more of the low-intensity probe sets are filtered out of the plot. All probe sets are ranked according to average signal level, and those in the lowest 25th (b), 50th (c), or 75th (d) percentile of signal level are eliminated from (a). Fit parameters: (b) R2 = 0.870; slope = 0.546; y-intercept = -0.008; (c) R2 = 0.895; slope = 0.517; y-intercept = -0.015; (d) R2 = 0.906; slope = 0.457; y-intercept = -0.017.
Choe et al. Genome Biology 2005 6:R16 doi:10.1186/gb-2005-6-2-r16