Table 3. F-values and corresponding degrees of freedom (DoF) of the three-way repeated measures ANOVA investigating the effects of the factors K (ratio between data samples and number of model parameters), SNR (ratio between the squared amplitude of the signal and the square amplitude of the noise) and TYPE (estimator used, i.e., OLS or ANN) on the performance parameters of GC estimation (BIAS0, BIAS1) and of network reconstruction (FNR, FPR, AUC). ***p < 10−5; ** 10−5 < p < 0.01.
| Factors | DoF | BIAS0 | BIAS1 | FNR | FPR | AUC |
|---|---|---|---|---|---|---|
| TYPE | (1, 99) | 5,901*** | 77.8*** | 68.8*** | 27.4*** | 36.9*** |
| SNR | (4, 396) | 328.1*** | 1,621.4*** | 645.1*** | 173.3*** | 761.2*** |
| K | (2, 198) | 9,785.3*** | 0.2 | 2,118.7*** | 10.2*** | 1881.1*** |
| TYPE × SNR | (4, 396) | 85.6*** | 199.7*** | 2.6** | 46.2*** | 10.5*** |
| TYPE × K | (2, 198) | 8,578*** | 99.9*** | 1,093*** | 280.8*** | 570.1*** |
| SNR × K | (8, 792) | 33.3*** | 167.4*** | 50.4*** | 19.4*** | 30.3*** |
| TYPE × K × SNR | (8, 792) | 26.1*** | 128.8*** | 65.4*** | 13.3*** | 45.4*** |