Journal of NeuroEngineering and Rehabilitation

(TFM, TFC, TT, Controls) and ALL are reported in Table 2 . R 2 was at least 0.70 for all amputees, with RMSE < 0.042. Therefore, the answer to Q2 was positive and only SPS was further considered. Question Q3 Figure 3a and b report the regression analysis for IMS vs SNS and P1S vs SNS, respectively, for ALL. R 2 and RMSE values for each group (TFM, TFC, TT, Controls) and ALL are reported in Table 2 . For IMS vs SNS, R 2 was lower than 0.64 for TT, with RMSE > 0.128. For P1S vs SNS, R 2 was lower than 0.2 for all amputees. Therefore, the answer to Q3 was negative and IMS, P1S and SPS were separately considered in all subsequent analyses. Questions Q4-Q6 Figures 4a , 5a and 6a report the distribution of SNS, IMS and P1S for TFC, TFC, TT and Controls. Numerical values are reported in Table 3 . For SNS and IMS, the Kruskal-Wallis test showed statistically significant differences among the medians of the groups ( p < 0.0001) (Figs. 4b and 5b ) . The pairwise analyses for: SNS (Fig. 4c ) showed that all amputee groups are different among each other, supporting a positive answer for Q4 and Q5 ; IMS (Fig. 5c ) showed a statistically significant difference between TFM and all other groups, with all TFM values > 1 as opposed to TFC and TT. This supports a partially positive answer to Q4 , a positive answer to Q5 and a negative answer to Q6. For P1S, the Kruskal-Wallis test reported a statistically significant difference in the medians among groups ( p = 0.0443) (Fig. 6b ) . The pairwise comparison did not show differences (Fig. 6c ) . This is a very possible situation for three reasons: the Kruskal-Wallis and pairwise comparisons try to negate different hypotheses; we applied a quite conservative multiple comparison strategy (HSD); the statistical power is reduced by the decreased number of transfemoral amputees (TF) within Two-Peaks . Fig. 2 Step symmetry index (SPS) vs Stance symmetry index (SNS). Each dot represents one subject. Subjects of the same group feature the same color (see legend in the plot). The purple parabolic line is the regression line for ALL subjects together. The equation of the fitting is reported on the right, with the fitting quality parameters R 2 (coefficient of determination) and RMSE (Root Mean Squared Error) Table 2 Quality of fit of the regressions for step (SPS), impulse (IMS) and first peak symmetry (P1S) indexes vs stance symmetry index (SNS) SPS vs SNS IMS vs SNS P1S vs SNS R 2 RMSE R 2 RMSE R 2 RMSE TFM 0,97 0,042 0,81 0,137 0,06 0,154 TFC 0,81 0,042 0,69 0,090 0,04 0,177 TT 0,70 0,036 0,37 0,128 0,20 0,289 CONTROLS 0,51 0,017 0,47 0,040 0,05 0,038 ALL 0,95 0,037 0,79 0,103 0,00 0,247 The coefficient of determination ( R 2 ), and the Root Mean Squared Error (RMSE) are reported for every group ( TFM transfemoral mechanical knee users, TFC transfemoral C-leg users, TT transtibial amputees, Controls), and for all subjects altogether (ALL). Bold: R 2 > 0.64, Regular: 0.36 < R 2 < 0.64 , Italic: R 2 < 0.36 [ 30 ] Cutti et al. Journal of NeuroEngineering and Rehabilitation 2018, 15 (Suppl 1):61 Page 33 of 72