Fig. 3

Out-of-sample prediction error for stutter peaks, reverse and forward. Markers indicate mean values, and bars extend to ± one standard deviation. k=10 out-of-sample prediction errors per dataset were estimated using cross validation. Over the12 datasets (6 reverse stutter, 6 forward stutter), we obtained 120 values for each model, with: (SR1) Stutter ratio model with parent peak height; (SP1) Peak height model with decayed amplitude; (SP2) Peak height model with parent peak height model; (SPE1 to SPE3) Exponential parent peak height models with n=5 to n=8 free parameters