Table 2.
Comparison of the Different Methods on the POPRES Data Set
| Algorithm | Euclidean Distance: 2nd[1st, 3rd] Quartile | Distance: 2nd[1st, 3rd] Quartile (km) | Relative Distance |
|---|---|---|---|
| PCA | 2.88 [1.68, 4.50] | 253.8 [150.0, 373.8] | 1.20 |
| PCA pruned | 2.78 [1.68, 4.61] | 247.2 [154.7, 378.4] | 1.17 |
| SMARTPCA | 2.88 [1.68, 4.49] | 254.2 [150.1, 373.8] | 1.20 |
| SMARTPCA pruned | 2.78 [1.67, 4.61] | 247.1 [155.0, 378.3] | 1.17 |
| SMARTPCA with regression (5) | 2.74 [1.62, 4.33] | 237.5 [146.4, 363.0] | 1.12 |
| SPA | 2.88 [1.65, 4.44] | 249.1 [148.4, 366.2] | 1.18 |
| SPA pruned | 2.55 [1.56, 4.02] | 226.4 [137.7, 336.2] | 1.07 |
| LOCO-LD | 2.42 [1.36, 3.70] | 211.2 [124.7, 313.8] | 1 |
“PCA” is our implementation of PCA. “Pruned” denotes running the methods on the data set after pruning for local and long-range LD. “SMARTPCA with regression (5)” denotes running SMARTPCA with the local regression option and setting the relevant parameter to 5. Reported error measures are the same as in Table 1. “Relative Distance” gives the ratio between the median error (in km) and LOCO-LD’s result.