Fig. 3. Distributions of the CPM prediction performance (as measured by the Spearman correlation between true and predicted values) for the RRS-B score change following NF training, compared to the baseline assessment.

Each point represents a single iteration from the 5-fold cross-validation results (with a total of 100 iterations using different random splits). The violin plot illustrates the distribution curve, and the horizontal lines indicate quartile positions. *: p < 0.05.