Journal of NeuroEngineering and Rehabilitation

Statistical analysis The distribution of the four indexes of symmetry (SPS, SNS, IMS and P1S) was checked for normality within each group (TFM, TFC, TT and Controls) and over all subjects, both visually with the Normal Probability Plot and with the Lilliefors test. This last failed for SPS TFM and P1S TT and there were doubts about IMS in general. The relationship between SNS and the three indexes SPS, IMS and P1S was evaluated with regression methods with the MATLAB Curve Fitting Toolbox. The strength of the relationship was primarily evaluated in terms of R 2 . This statistical parameter, multiplied by 100, is usually interpreted as the variance of “ y ” accounted for by “ x ” , where in this case “ y ” is SPS or IMS or P1S, and “ x ” is SNS. In addition, the root-mean-square error (RMSE) of the residuals was also reported. Distributions were reported in terms of median and interquartile range [ 3 ] , with box plots. For each symmetry index, the Kruskal-Wallis test ( α = 0.05) was adopted to check for overall statistically significant differences among TFM, TFC, TT and Controls. In identifying pairwise differ- ences, the Tukey-Kramer “ HSD ” correction was applied within the MATLAB “ multcompare ” function. Results Gait speed was compared among TFM (1.12 ± 0.13 m/s), TFC (1.17 ± 0.12 m/s), TT (1.23 ± 0.19 m/s) and Controls (1.41 ± 0.21 m/s). ANOVA did not show statistically significant differences among amputees ( p = 0.14), but only between Controls and amputees ( p = 0.0005). Further results are reported hereinafter based on their relevance for questions Q1-Q6. Question Q1 Figure 1 reports the number of subjects in subgroups Two-Peaks , which decreases from TT (20/23), to TFM (7/12) to TFC (10/25). The number of TFC with non- standard GRF is remarkably high (60%); these patients report a consistent “ alternative ” pattern (example provided in Fig. 1b ) . Based on these results, the answer to Q1 was negative and the calculation of the symmetry index P1S was restricted to the subjects in Two-Peaks . Question Q2 Figure 2 reports the regression analysis for SPS vs SNS considering the whole set of patients and Controls ( “ ALL ” in brief ). R 2 and RMSE values for each group Fig. 1 a Number of subjects in subgroup Two-Peaks for TFM (transfemoral mechanical knee users), TFC (transfemoral C-leg users), TT (transtibial amputees), and Controls: b typical alternative vertical ground reaction force pattern shown by TFC patients not included in Two-Peaks Cutti et al. Journal of NeuroEngineering and Rehabilitation 2018, 15 (Suppl 1):61 Page 32 of 72