Table 3.
Results of K = 100 cross-validation experiments with the HDRS item 1 score (left columns) and YMRS item 1 score (right columns) models based on all, mandatory and mood self-assessment items, respectively.
| HDRS item 1 score | YMRS item 1 score | |||
|---|---|---|---|---|
| Model | R2 (SD) ↑a | RMSE (SD) ↓b | R2 (SD) ↑a | RMSE (SD) ↓b |
| All self-assessment items | ||||
| Pooled naïve mean | −0.03 (0.04) | 0.95 (0.07) | −0.05 (0.07) | 0.61 (0.10) |
| Pooled Ridge | 0.41 (0.08) | 0.71 (0.06) | −0.09 (0.13) | 0.62 (0.09) |
| Pooled XGBoost | 0.50 (0.11) | 0.66 (0.07) | −0.17 (0.20) | 0.64 (0.09) |
| Pooled Bayesian | 0.38 (0.14) | 0.73 (0.10) | −0.16 (0.20) | 0.63 (0.09) |
| Separate naïve mean | 0.35 (0.15) | 0.75 (0.08) | −0.27 (0.33) | 0.66 (0.08) |
| Separate Ridge | 0.37 (0.15) | 0.73 (0.07) | −0.23 (0.30) | 0.65 (0.08) |
| Separate XGBoost | 0.18 (0.17) | 0.84 (0.07) | −0.35 (0.34) | 0.68 (0.06) |
| Hierarchical Bayesian | 0.40 (0.12) | 0.72 (0.06) | −0.07 (0.24) | 0.61 (0.08) |
| Mandatory self-assessment items | ||||
| Pooled naïve mean | −0.03 (0.04) | 0.93 (0.06) | −0.04 (0.06) | 0.60 (0.08) |
| Pooled Ridge | 0.32 (0.07) | 0.75 (0.05) | 0.01 (0.10) | 0.59 (0.08) |
| Pooled XGBoost | 0.39 (0.13) | 0.71 (0.07) | −0.17 (0.22) | 0.63 (0.08) |
| Pooled Bayesian | 0.33 (0.13) | 0.75 (0.09) | −0.03 (0.17) | 0.59 (0.08) |
| Separate naïve mean | 0.35 (0.13) | 0.73 (0.08) | −0.25 (0.27) | 0.65 (0.08) |
| Separate Ridge | 0.37 (0.13) | 0.72 (0.08) | −0.22 (0.25) | 0.64 (0.08) |
| Separate XGBoost | 0.14 (0.14) | 0.84 (0.07) | −0.36 (0.36) | 0.67 (0.06) |
| Hierarchical Bayesian | 0.44 (0.10) | 0.68 (0.07) | 0.00 (0.22) | 0.58 (0.08) |
| Mood self-assessment item | ||||
| Pooled naïve mean | −0.03 (0.04) | 0.94 (0.07) | −0.07 (0.15) | 0.61 (0.09) |
| Pooled Ridge | 0.34 (0.07) | 0.75 (0.05) | 0.01 (0.16) | 0.58 (0.09) |
| Pooled XGBoost | 0.40 (0.12) | 0.71 (0.07) | −0.04 (0.27) | 0.59 (0.09) |
| Pooled Bayesian | 0.33 (0.12) | 0.76 (0.09) | 0.02 (0.21) | 0.58 (0.08) |
| Separate naïve mean | 0.34 (0.12) | 0.75 (0.07) | −0.36 (0.62) | 0.66 (0.08) |
| Separate Ridge | 0.36 (0.12) | 0.74 (0.07) | −0.35 (0.61) | 0.66 (0.08) |
| Separate XGBoost | 0.37 (0.13) | 0.73 (0.07) | −0.21 (0.51) | 0.63 (0.08) |
| Hierarchical Bayesian | 0.47 (0.10) | 0.67 (0.07) | −0.08 (0.45) | 0.59 (0.08) |
The best HDRS item 1 result was achieved using the XGBoost model with all self-assessment items while YMRS item 1 could not be estimated significantly better than the naïve baseline models.
Bold values indicates the best results within each set of self-assessment items.
aCoefficient of determination. Higher is better.
bRoot Mean Square Error. Lower is better.