Table 3.
Comparison of sample size calculation methods, including the proposed method in this paper, Zhao et al.’s approach (RnaSeqSampleSize) and Wu et al.’s approach (PROPER).
| Simulation 1 | Simulation 2 | Simulation 3 | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Method | Sample | Power | Computation | Sample | Power | Computation | Sample | Power | Computation |
| size | time | size | time | size | time | ||||
| Our method | 34 | 0.815 | 17.3 sec | 13 | 0.823 | 16.8 sec | 22 | 0.804 | 15.5 sec |
| RnaSeqSampleSize | 32 | 0.810 | 0.3 sec | 9 | 0.765 | 54.6 sec | 74 | 0.742 | 82.8 sec |
| PROPER | 25 | 0.806 | 6.5 h | 10 | 0.804 | 1.5 h | 19 | 0.805 | 3.5 h |
Results determined by our method were based on parameters estimated at m=200. Power was evaluated based on the voom and limma pipeline for our method, while edgeR for RnaSeqSampleSize and PROPER. The computation time for each simulation was calculated on a MacAir laptop with 1.3 GHz i7 CPU and 4GB RAM