Table 1.

Rank methods based on different measures of forecast accuracy

Threshold APE, %
Overall rankingKS50809095MeanRMSE
1. Myrskylä et al. (16)17 (1)0.8 (1)4.5 (3)8.3 (4)12.5 (5)2.9 (4)5.9 (5)
2. Schmertmann et al. (20)16 (2)0.9 (2)3.9 (1)7.1 (2)11.5 (3)2.7 (2)5.4 (3)
3. de Beer (12, 13)16 (2)1.0 (3)5.0 (4)8.3 (3)11.4 (2)2.9 (3)5.3 (2)
4. Ševcíková et al. (19)15 (4)1.1 (5)4.0 (2)6.6 (1)9.5 (1)2.4 (1)4.2 (1)
5. Freeze rates11 (5)1.6 (9)5.8 (7)8.7 (5)11.5 (4)3.3 (5)5.5 (4)
6. Lee (14)11 (5)1.2 (7)6.1 (8)10.8 (7)15.1 (7)3.6 (6)6.9 (6)
7. Evans (26)11 (5)1.2 (6)5.6 (5)10.7 (6)16.0 (8)3.8 (7)7.7 (8)
8. Hyndman and Ullah (15)11 (5)1.1 (4)6.3 (9)11.8 (9)17.5 (9)3.9 (8)7.8 (9)
9. Brass (8)8 (9)2.4 (11)7.8 (10)11.7 (8)14.7 (6)4.5 (9)7.1 (7)
10. Cheng and Lin (10)8 (9)1.3 (8)5.8 (6)14.0 (10)23.5 (12)4.9 (10)11.2 (12)
11. Li and Wu (22)8 (9)2.5 (12)8.9 (11)14.1 (11)21.8 (11)5.7 (11)10.7 (11)
12. Myrskylä and Goldstein (27)5 (12)2.0 (10)14.4 (15)28.2 (15)38.8 (13)8.5 (13)16.2 (13)
13. Willekens and Baydar (9)4 (13)2.7 (14)14.9 (16)28.2 (16)41.9 (15)9.5 (14)19.6 (15)
14. Chandola et al. (5)4 (13)2.7 (15)13.0 (14)26.5 (14)41.6 (14)9.6 (15)24.4 (18)
15. Peristera and Kostaki (4)4 (13)2.5 (13)11.9 (12)24.0 (13)42.3 (16)10.3 (16)29.8 (19)
16. Schmertmann (3)0 (16)6.8 (19)12.3 (13)16.2 (12)21.6 (10)8.2 (12)10.6 (10)
17. Coale and McNeil (6)0 (16)4.9 (17)16.4 (17)30.9 (17)43.7 (17)10.8 (17)18.4 (14)
18. Hadwiger (2)0 (16)3.9 (16)18.0 (18)39.6 (19)57.6 (20)12.0 (18)21.9 (16)
19. Peristera and Kostaki (4)0 (16)6.2 (18)19.9 (19)34.1 (18)55.8 (19)15.0 (19)33.0 (20)
20. Coale and Trussell (7)0 (16)7.2 (20)29.7 (20)42.8 (20)52.9 (18)15.1 (20)23.1 (17)
No. of inversions017171721924
  • The overall ranking of methods in column 1 is based on the KS test statistic of stochastic dominance (column 2) and the overall mean (column 7). Testing data are from the HFD (details are in SI Appendix, section 2).