Methods: | fPH | NN | WNN | SVM | NeuN | MC | BL | Npute |
---|
0.01-human-0.5% | 860.44 | 0.22 | 0.28 | 156.09 | 13.08 | 0.10 | 0.07 | - |
0.01-mouse-0.5% | 530.59 | 0.17 | 0.16 | 33.47 | 0.43 | 0.12 | 0.08 | 21.08 |
hd-cattle-0.5% | 6178.95 | 0.37 | 0.41 | 931.16 | 43.79 | 0.22 | 0.16 | - |
ld-cattle-0.5% | 36209.82 | 1.15 | 2.41 | 22570.29 | 847.00 | 0.54 | 0.18 | - |
- Average imputation runtime of all methods on four datasets: 0.01-human-0.5%, 0.01-mouse-0.5%, hd-cattle-0.5%, ld-cattle-0.5%, where the average is taken over 10 simulated datasets (with 5 runs using different genetic distance thresholds, except fastPHASE, Npute, BaseLine). ‘fPH, NeuN’ stand for ‘fastPHASE, NeuralNet’, respectively. Note that Npute spent most of its time, 20.62 out of the 21.08 seconds, in training for selecting the best window size in the range [1,50].