Pre and Post-filter CNV counts.

The table show the CNV-counts for Cases and Controls combined (n = 20,100) at four levels of filtering. First, Raw PennCNV reflect total CNV-counts initially identified by PennCNV. The second level, PennCNV 10-SNP, show the counts after applying a minimum 10-SNP window. Next, the Post-filter counts are shown, which reflect the counts after a series of empirically-derived CNV-quality filters were applied. The empirically derived criteria were set after a meticulous review of a large number of the candidate CNVs included in PennCNV 10-SNP. After grouping the Post-filter CNVs by sample it became evident that a small subset of samples (2–3%) had very high CNV-counts, and visual inspection revealed a majority of the CNVs in these samples to be false. To eliminate the excessive CNV-counts an outlier-filter was applied. The CNVs remaining after Outlier-filter were used in the association analysis. The average CNV-counts per individual are shown in parenthesis. The right-most column show the False Positive Rate (FPR) determined by visual inspection of 1,000 CNVs randomly selected at each step.