Table 2 Comparison of the average classification test error (%) in 5-fold cross-validation using different number of top genes selected by: (1) a set of methods without adjustment for heterogeneity: PLR, SOS, IndAss; (2) another set of methods accounting for heterogeneity: SVA and SOSA respectively in the Pan-Cancer study
From: An embedded method for gene identification problems involving unwanted data heterogeneity
| Â | Â | T10 | T30 | T50 | T100 | T200 | T300 | T400 |
|---|---|---|---|---|---|---|---|---|
Unadjusted | PLR | 1.9 | 0.4 | 0.4 | 0.4 | 0.1 | 0.1 | 0.1 |
| Â | SOS | 3.1 | 0.4 | 0.3 | 0.3 | 0.3 | 0.3 | 0.3 |
| Â | IndAss | 4.5 | 1.5 | 1.8 | 0.1 | 0.3 | 0.1 | 0.1 |
Adjusted | SVA | 4.3 | 1.5 | 1.3 | 0.6 | 0.3 | 0.3 | 0.1 |
| Â | SOSA | 1.1 | 0.5 | 0.3 | 0.1 | 0.1 | 0.1 | 0.1 |
Baseline | Using all 20,531 genes | 0.5 | Â | Â | Â | Â | Â | Â |