*4.6. Statistical Analysis*

Descriptive statistical analysis of the collected data by Mann-Whitney U and χ<sup>2</sup> tests did not show significant differences among PD and HC (age, gender, cognitive status), then the data were safely pooled into two groups (PD and HC) for the following analyses.

We note that, as common in feature selection algorithms, the previous selection of parameters based on the EN algorithm assumes the UPDRS scores are ratio data type, while they actually are ordinal data. Then, to confirm the relevance of these parameters in the context of the ordinal nature of the scores and to deal with their non-normal distributions, Spearman non-parametric rank correlation at a significant level *p* < 0.05 was applied. Only those parameters showing a Spearman's correlation coefficient ρ greater than 0.3 (as absolute value) with respect to the UPDRS scores assigned to the LA, AC and Po task performances were considered for the final sets. For the PSCOM task, the CoM parameters were correlated with the PIGD subscale scores (PSPIGD).

Furthermore, as a support to the effectiveness of the selected parameters in the automated assessment, the statistical significance of each parameter in discriminating PD and HC was considered and verified by the Mann-Whitney U test with *p* < 0.05. All statistical analyses were performed using Matlab. For the correct application of the test, only the data of the second acquisition session were considered.
