Table 6.
Classification/prediction error (SD), p = 1000
| Scenario 1 | Scenario 2 | Scenario 3 | Scenario 4 | |||
|---|---|---|---|---|---|---|
| RF | 24.1% (2.7%) | 10.01 (1.32) | 9.13 (0.52) | 7.09 (0.65) | ||
|
|
22.7% (3.0%) | 8.71 (1.38) | 8.48 (0.85) | 5.42 (0.55) | ||
| RF-log(p) | 19.3% (3.7%) | 5.89 (2.40) | 8.93 (1.37) | 3.50 (0.35) | ||
| ET | 21.6% (4.0%) | 6.60 (1.65) | 9.09 (0.51) | 5.38 (0.51) | ||
| BART | 30.0% (6.2%) | 9.88 (1.30) | 9.14 (0.54) | 3.77 (0.52) | ||
| Lasso | 26.6% (3.6%) | 10.27 (1.19) | 9.07 (0.58) | 1.15 (0.09) | ||
| Boosting | 24.8% (3.1%) | 9.78 (1.16) | 9.05 (0.54) | 3.22 (0.37) | ||
| RLT-naive | 22.4% (2.5%) | 6.70 (1.90) | 9.01 (0.64) | 5.39 (0.58) | ||
|
| ||||||
| RLT
|
||||||
| Muting | Linear combination | |||||
|
|
||||||
| No | 1 | 18.8% (4.4%) | 5.64 (1.51) | 7.81 (1.07) | 5.08 (0.60) | |
| 2 | 21.0% (4.0%) | 6.97 (1.58) | 7.84 (0.87) | 3.47 (0.52) | ||
| 5 | 23.6% (3.4%) | 7.66 (1.57) | 8.01 (0.89) | 3.39 (0.52) | ||
| Moderate | 1 | 16.0% (5.0%) | 4.50 (1.47) | 7.48 (1.26) | 3.81 (0.45) | |
| 2 | 17.5% (4.5%) | 5.45 (1.68) | 7.48 (1.06) | 2.60 (0.39) | ||
| 5 | 20.4% (4.0%) | 6.26 (1.73) | 7.60 (0.98) | 2.49 (0.39) | ||
| Aggressive | 1 | 13.7% (4.9%) | 4.01 (1.38) | 7.20 (1.22) | 3.36 (0.42) | |
| 2 | 14.2% (5.1%) | 4.24 (1.55) | 7.07 (1.16) | 2.03 (0.29) | ||
| 5 | 16.1% (4.8%) | 5.05 (1.73) | 7.09 (1.05) | 1.91 (0.29) | ||
For each scenario, the best two methods within each panel are bolded. The overall best method is underlined.