Fig. 7

Objective comparison of threshold finding methods using evaluation curves. Four methods are compared: manual thresholds (blue, dotted lines), SLR estimations (red, dashed lines), NN predictions (green, dash-dotted lines), and an “always 50 dB” control method (grey, solid lines). Separate plots show evaluation curves for each stimulus (click, 6, 12 18, 24, 30 kHz). Plots show the normalized time variance of the averaged signal \(S^2(n)/S^2(N)\) (y-axis) vs. the total percentage of ABR curves included in the cumulative average n/N (x-axis). a shows NN predictions and SLR estimations from experiments 1 and 5, b shows NN predictions and SLR estimations from experiments 4 and 8, as introduced in Table . Two methods can be compared in a way that the curve of the better method stays longer close to zero