Table 2.
Factors contributing to performance of pathway-based RSF prediction models. Shown are test set errors using 1-C, averaged over 10 independent experiments.
Method | Miller et al. (2005) data | Loi et al. (2008) data |
---|---|---|
Pathway-based RSF | 0.2823 | 0.2881 |
(1) No gene screening in gene sets | 0.2992 | 0.3014 |
(2) No K-means clusters | 0.2903 | 0.2934 |
(3) No clinical variables | 0.2895 | 0.2925 |