Statistical analysis

Using the Kolmogorov–Smirnov test it appeared that most outcome measures for the pre-test had non-normal distributions or were close to non-normal distribution. (Attempts needed for successful switch D(28) = 0.18, p = 0.021, Time till switch D(28) = 0.16, p = 0.067, Switching Accuracy D(28) = 0.31, p = 0.000). Therefor non-parametric tests were chosen for analyses.

To check for initial differences between groups, a non-parametric Kruskal–Wallis test for independent variables was applied to the pre-test data of the clinically relevant performance measures.

Subsequently transfer effects from training to prosthesis use (aim 1) were determined by analysing differences between pre- and post-test, using a non-parametric Friedman test for dependent variables (within group comparison). Differences in transfer effects induced by type of training programs (aim 2) were determined by using a non-parametric Kruskal–Wallis test for independent variables. This test was applied to the difference scores between the pre-test and post-test data (between group comparison).

To provide insight into changes in clinical and myosignal features of the sEMG during the 5-day training program (aim 3) the Pearson correlation coefficient (r) of the least square line of the outcomes per day across the 5 training days was computed for all outcome measures. For the outcome measures Accuracy and Amplitude a positive value of r indicated improvement. For the outcome measures Attempts, Time needed to switch, Phasing and Width a negative r value indicated improvement. Furthermore since we expected that over training the variability of the sEMG became smaller, a non-parametric Friedman test was used to see whether the interquartile ranges (IQRs) of the myosignal features changed over time. The IQR has been calculated per individual and those values are used in the Friedman test. This test thus analyses whether the median value of the within participant IQR differs. For all tests P-values smaller than 0.05 were considered to be statistically significant. If significant changes were found a post-hoc Wilcoxon signed-rank test with Bonferroni correction was performed to determine which training session(s) was/were significantly different from session 1.