Matlab statistics and machine learning toolbox

Paired samples tests were carried out within MATLAB Statistics and Machine Learning Toolbox (MathWorks, USA) to test for significant changes following surgery. Where parametric assumptions were not met, a Wilcoxon signed rank test was used. A t-test was used to identify differences between the post-operative TKR and the NP group. Where parametric assumptions were not met, the Mann-Whitney test was used. A Bonferroni correction was used to adjust for multiple comparisons. All statistical inferences were calculated using the MATLAB Statistics and Machine Learning Toolbox (MathWorks, USA).

Thrombin dynamics analysis was performed in Matlab (R2016a, Mathworks, Eindhoven, the Netherlands) using the Matlab “Statistics and Machine Learning Tool Box” (version 10.2, Mathworks, Eindhoven), and theMatlab “Curve Fitting Toolbox” (version 3.5.3, Mathworks, Eindhoven, the Netherlands) and to ensure a high throughput computation of thrombin dynamics parameters. Analyses were performed on a HP probook (HP, Amstelveen, the Netherlands) equipped with a Microsoft Windows 10 pro operating system. The equations described in the sections above were implemented in an semi-automized data analysis script, which analyzes around 50 curves per minute.

Spearman’s correlation coefficients were used to compare clinical scores and RT parameters (corrected for age, sex, handedness and direction; see ‘Data analysis’). Significant difference between two groups of participants (e.g. individuals with RHD and LHD, and male and female) was performed by comparing the two distributions with a Kolmogorov–Smirnoff test. Some of our analyses were performed on circular data, and, as such, commonly used methods for determining correlations are not suitable. We used Rayleigh’s test^{38 ,39} to identify whether the observed RTs were unimodally ‘tuned’ across the eight spatial directions with slower RTs in one spatial direction and faster RTs in the opposite direction. Finally, we used the Watson–Williams test to determine whether the directional tuning was different between two sets of circular data (i.e. RT directional tuning for left and right arms across all individuals with stroke). The Watson–Williams test was only reported if the data passed the assumptions of the test. All statistical calculations were performed with MATLAB Statistics and Machine Learning Toolbox (Mathworks, R2018b) and the toolbox for circular statistics with MATLAB.³⁹