Table 1

Errors of the simplest machine learning model

Model (n = 96)
Error +/- S.E.

Cross-validation
0.33
Random guessing (n = 10)
0.93 +/- 0.01
Re-permuted and cross-validated (n = 10)
0.94 +/- 0.01

We used leave-one-out cross-validation of all samples simultaneously. With random guessing we compared the matching of the true order of classification labels against a randomly generated set, which produces a baseline for cross-validation comparison. For re-permutation, the true labels were randomized to create a semi-random false dataset.

Henderson et al. Genome Biology 2005 6:R76   doi:10.1186/gb-2005-6-9-r76