Table 4.
Classification/prediction error (SD), p = 200
| Scenario 1 | Scenario 2 | Scenario 3 | Scenario 4 | |||
|---|---|---|---|---|---|---|
| RF | 21.3% (3.5%) | 8.35 (1.28) | 8.66 (0.55) | 5.93 (0.61) | ||
|
|
19.8% (3.6%) | 6.50 (1.30) | 6.97 (0.88) | 4.35 (0.47) | ||
| RF-log(p) | 15.2% (3.3%) | 4.55 (1.23) | 7.75 (1.74) | 3.23 (0.33) | ||
| ET | 18.3% (4.2%) | 4.61 (1.26) | 8.26 (0.60) | 4.57 (0.51) | ||
| BART | 25.7% (2.8%) | 8.00 (1.13) | 8.13 (0.83) | 2.63 (0.30) | ||
| Lasso | 26.5% (2.6%) | 9.99 (1.02) | 8.96 (0.50) | 1.12 (0.07) | ||
| Boosting | 21.3% (2.8%) | 8.47 (0.97) | 8.60 (0.53) | 2.85 (0.35) | ||
| RLT-naive | 19.0% (4.3%) | 4.65 (1.51) | 7.77 (0.69) | 4.59 (0.54) | ||
|
| ||||||
| RLT
|
||||||
| Muting | Linear combination | |||||
|
|
||||||
| None | 1 | 14.8% (4.0%) | 4.09 (1.00) | 5.43 (0.75) | 4.36 (0.52) | |
| 2 | 16.5% (4.3%) | 4.93 (1.27) | 5.71 (0.65) | 2.88 (0.44) | ||
| 5 | 18.9% (4.2%) | 5.52 (1.43) | 5.85 (0.62) | 2.80 (0.44) | ||
| Moderate | 1 | 11.8% (3.4%) | 3.20 (0.84) | 4.80 (0.74) | 3.27 (0.39) | |
| 2 | 12.2% (3.6%) | 3.43 (0.96) | 4.85 (0.71) | 2.13 (0.30) | ||
| 5 | 14.2% (4.0%) | 3.90 (1.18) | 4.89 (0.69) | 2.03 (0.30) | ||
| Aggressive | 1 | 10.3% (3.2%) | 2.79 (0.71) | 4.87 (0.81) | 3.23 (0.39) | |
| 2 | 9.8% (3.1%) | 2.66 (0.76) | 4.90 (0.80) | 1.84 (0.23) | ||
| 5 | 11.1% (3.4%) | 2.95 (0.94) | 4.74 (0.82) | 1.71 (0.22) | ||
For each scenario, the best two methods within each panel are bolded. The overall best method is underlined.