Table 4. Performance of Classifiers for Psoriasis data.
| A. Binary Classifiers | |||||||
| Training (n = 360) | Test set (n = 89) | ||||||
| Method | # genes | Error (%) | 5-fold CV (%) | GBS | PE (%) | GBS | |
| LS vs Normal Training: 233 Test: 49 | TDGR (adjusted) | 30 | 0 | 0.86 | 0.0001 | 0 | 0.0006 |
| Meta-TGDR (unadjusted) | 18 | 0 | 0.86 | 0.0010 | 2.04 | 0.0084 | |
| Meta-TGDR (adjusted) | 22 | 0 | 0.86 | 0.0006 | 0 | 0.0028 | |
| TDGR w/Bagging (adjusted, BF >30%) | 18 | 0 | – | 0.0011 | 0 | 0.0004 | |
| Meta-TGDR w/Bagging (adjusted, BF >30%) | 10 | 0 | – | 0.0012 | 0 | 0.0032 | |
| LS vs NL Training: 271 Test: 68 | TDGR (adjusted) | 35 | 0 | 1.48 | 0.0009 | 1.47 | 0.0136 |
| Meta-TGDR (unadjusted) | 26 | 1.11 | 1.85 | 0.0105 | 2.94 | 0.0294 | |
| Meta-TGDR (adjusted) | 25 | 0 | 1.48 | 0.0036 | 1.47 | 0.0143 | |
| TDGR w/Bagging (adjusted, BF >30%) | 22 | 0 | – | 0.0021 | 1.47 | 0.0144 | |
| Meta-TGDR w/Bagging (adjusted, BF >40%) | 16 | 1.48 | – | 0.0041 | 1.47 | 0.0142 | |
| NL vs Normal Training: 216 Test: 61 | TDGR (adjusted) | 26 | 0 | 0 | 1.5×10−5 | 0 | 7.3×10−5 |
| Meta-TGDR (unadjusted) | 40 | 5.56 | 18.06 | 0.0570 | 8.20 | 0.0659 | |
| Meta-TGDR (adjusted) | 22 | 0 | 1.85 | 0.0032 | 0 | 0.0054 | |
| TDGR w/Bagging (adjusted, BF >30%) | 24 | 0 | – | 2.4×10−5 | 0 | 7.3×10-5 | |
| Meta-TGDR w/Bagging (adjusted, BF >40%) | 21 | 0 | – | 0.0033 | 0 | 0.0054 | |
A. Comparison between TGDR and Meta-TGDR for binary classifiers. B. Comparisons between TGDR and Meta-TGDR for 3-class classifiers.