Table 4.
Mean rank and statistical significance from post hoc Nemenyi test. For each season-state combination, the model forms are ranked from best (rank=1) to worst (rank=8).
|
Measure |
Model |
Mean rank |
GFTa |
RRO |
RRR |
RRS |
SS0 |
SRR |
SRS |
| Pearson correlation coefficient (COR) | GFT | 2.67 | |||||||
| RR0 | 4.55 | <.001 | |||||||
| RRR | 3.34 | .002 | <.001 | ||||||
| RRS | 3.68 | <.001 | <.001 | .68 | |||||
| SS0 | 6.87 | <.001 | <.001 | <.001 | <.001 | ||||
| SRR | 4.37 | <.001 | .98 | <.001 | .01 | <.001 | |||
| SRS | 4.75 | <.001 | .97 | <.001 | <.001 | <.001 | .55 | ||
| SSS | 5.73 | <.001 | <.001 | <.001 | <.001 | <.001 | <.001 | <.001 | |
| Root mean square error (RMSE) | GFT | 4.46 | |||||||
| RR0 | 5.27 | .002 | |||||||
| RRR | 4.68 | .96 | .06 | ||||||
| RRS | 4.82 | .62 | .35 | .99 | |||||
| SS0 | 5.77 | <.001 | .19 | <.001 | <.001 | ||||
| SRR | 3.34 | <.001 | <.001 | <.001 | <.001 | <.001 | |||
| SRS | 3.71 | .005 | <.001 | <.001 | <.001 | <.001 | .61 | ||
| SSS | 3.96 | .2 | <.001 | <.001 | <.001 | <.001 | .04 | .92 | |
| Mean absolute proportion error (MAPE) | GFT | 5.26 | |||||||
| RR0 | 4.91 | .65 | |||||||
| RRR | 5.18 | .99 | .89 | ||||||
| RRS | 5.7 | .37 | .002 | .15 | |||||
| SS0 | 3.75 | <.001 | <.001 | <.001 | <.001 | ||||
| SRR | 3.17 | <.001 | <.001 | <.001 | <.001 | .07 | |||
| SRS | 3.93 | <.001 | <.001 | <.001 | <.001 | .99 | <.001 | ||
| SSS | 4.09 | <.001 | .001 | <.001 | <.001 | .69 | <.001 | .99 |
aGFT: Google Flu Trends.