Simulations: per-read error rate estimates based on simulated reads under five different error rates. The horizontal lines are drawn at 0, where the error rate is estimated perfectly. Each boxplot contains the differences between error rate estimates and the true error rate from 100 simulated data sets. Shadow regression and mismatch counting estimates are plotted next to each other for a given error rate.
Victoria Wang et al. BMC Bioinformatics 2012 13:185 doi:10.1186/1471-2105-13-185