Fig. 12

Comparison of mutant vs. control effect sizes between three hearing threshold finding methods. For two genes (Top: Hunk, Bottom: Ngdn), hearing thresholds determined by three different methods (manual, NN, SLR) are compared using conventional plots (panels A–C, E–G) and a new effect size plot (panels D, H). In the conventional plots, stimulus-specific hearing thresholds are compiled to median hearing curves of mutants (triangles and solid lines) and controls (squares and dashed lines), which are connected for the non-click stimuli. Whiskers indicate inter-quartile ranges. Numbers of mice in each group are shown in the legend. On the y-axis, hearing threshold is given in 20 dB ticks. The x-axis shows the stimulus. Effect size plots (D, H) show non-parametric effect size (Cliff’s delta [39], y-axis) of mutant vs. control animals for all six stimuli (x-axis). Colours indicate the threshold finding method (manual: blue, NN: red, SLR: orange). The grey dashdotted horizontal lines show the thresholds for strong effects at ±0.474. The central dotted horizontal line indicates the zero effect size level. Lines connect effect sizes of the same method. Top: For Hunk, effect sizes differ only marginally for all stimuli, consistent with mutant and control curves running close together in the conventional plots A–C. Bottom: For Ngdn, smaller effects are found with NN and SLR consistently throughout all stimuli, whereas manual thresholds show large effects. Again, this is consistent with the conventional plots (E–G), where mutant and control curves are clearly apart for manual thresholds (E), but close together for NN (F) and SLR (G). Interpretation aid: the effect size relates the difference between mutant and control medians to data variance (indicated by the interquartile range whiskers)