Table 2 Scenario 2: Single binary incomplete variable missing at random, 9 continuous complete variables, 100 random runs, 30% missing rate
From: Secure distributed multiple imputation enables missing data inference for private data proprietors
Final analysis | Imputation | Performance | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
Technology | Θ bias | Θ SD | Θ rMSE | \(| \widehat{{\rm{y}}}-{\rm{y}}|\) (μ) | \(| \widehat{{\rm{y}}}-{\rm{y}}|\) (σ) | Discr. | Accuracy | AUC | Time (s) | Net (MB) | |
500 inds. | Python | 0.268 | 5.09 × 10-16 | 0.268 | 0.141 | 0.092 | 0 | 0.466 | 0.477 | 0.063 | N/A |
PyMICE | 0.035 | 4.00 × 10-16 | 0.035 | 0.040 | 0.031 | N/A | 0.460 | 0.500 | 0.225 | N/A | |
SMC-MICE | 0.052 | 3.12 × 10-10 | 0.052 | 0.049 | 0.038 | 0 | 0.466 | 0.432 | 0.335 | 14.019 | |
MHE-MICE | 0.053 | 2.91 × 10-7 | 0.053 | 0.050 | 0.038 | 0 | 0.466 | 0.432 | 1753 | 17,061 | |
5000 inds. | Python | 0.263 | 4.71 × 10-16 | 0.263 | 0.134 | 0.078 | 0 | 0.508 | 0.509 | 0.143 | N/A |
PyMICE | 0.011 | 4.84 × 10-16 | 0.011 | 0.024 | 0.025 | N/A | 0.510 | 0.496 | 0.212 | N/A | |
SMC-MICE | 0.012 | 3.30 × 10-10 | 0.012 | 0.026 | 0.026 | 0 | 0.508 | 0.475 | 3.052 | 137.398 | |
MHE-MICE | 0.012 | 8.31 × 10-7 | 0.012 | 0.026 | 0.026 | 0 | 0.508 | 0.475 | 2321 | 27,038 | |
