Table 4.
Feature selection for simulated dataset 1 using Ordinary LR with FFS algorithm and 5-site EXPLORER with DFFS protocol, where β′ and β″ are model parameters averaged over 30 trials learnt by Ordinary LR and EXPLORER, respectively, the Prob. indicates the chance of a given covariate to be selected by either FFS or DFFS during the 30 trials, and two-sample Z-tests are performed between β′ and β″, which show no differences at all.
True value | Ordinary LR | EXPLORER | two-sample Z-test | ||||
---|---|---|---|---|---|---|---|
β′ | Prob. | β″ | Prob. | Test statistic | p-value | ||
β1 | −2.2428 | −2.0743 | 1.000 | −2.0722 | 1.000 | −0.3573 | 0.7209 |
β2 | 0.0000 | 0.0452 | 0.267 | 0.0527 | 0.300 | −0.3669 | 0.7137 |
β3 | 3.1668 | 3.0195 | 1.000 | 3.0167 | 1.000 | 0.5919 | 0.5539 |
β4 | 0.0000 | −0.0402 | 0.200 | −0.0265 | 0.133 | −0.7003 | 0.4837 |
β5 | −1.8701 | −1.7313 | 1.000 | −1.7316 | 1.000 | 0.1305 | 0.8962 |