Reproducibility of detections between independent subsamples. Reproducibility of PECA-SI, RMA-SI, PLIER-SI, RMA-MIDAS, PLIER-MIDAS, RM-LM, PLIER-LM and FIRMA in detecting differential splicing in the colon cancer data. The ability of the methods to detect the same candidate splicing events in independent subsamples was investigated at sample sizes of two to four (x-axis). The reproducibility was measured as the overlap of the top-ranked 500 detections between the subsamples. At each sample size, the average reproducibility over at least 15 randomly sampled pairs of datasets is shown together with the standard error of the mean (error bar). The same datasets were analyzed with the different detection methods. Reproducibility in random data is shown as a reference (0.002). Similar results were produced with the top-ranked 100, 1,000, 1,500 and 2,000 detections [see Additional data file 3].