Median-referenced classification results for HMM, GNB and null models.

The values in each cell represent the metric indicated by the column header. Precision, Recall and F1 score are computed using a macro average with uniform weights for each class. The values in parenthesis indicate the p-values associated with the random and mode null models respectively. When p < 0.05, the corresponding model performs better than the null model and that improvement is statistically significant at the 5% level.