We perform the calibration using a re-sampling data-driven strategy to generate pseudo null datasets (where we know that there is no copy number variation) on which we can apply our algorithms (details given in the Materials and Methods section). The rate at which copy number variant events were detected on these simulated null datasets then provides an empirical measure of the false positive rate for different values of L and BFthresh. In addition, we have similarly estimated the power of our procedure to detect events of various sizes at different false positive rates, by re-sampling BeadArray™ data from chromosome X multiple copy cell lines with known copy numbers 1–4.
