Table 3
Goodness-of-fit statistics for version 1 of the models. The goodness-of-fit statistics chosen are the root mean square error (RMSE) on a relative scale (rRMSE, RMSE divided by the median of the observed values), the median symmetric accuracy (MSA), the mean error (ME) on a relative scale (rME, ME divided by the median of the observed values), the precision and the correlation.
Model |
Goodness of fit |
|||||
---|---|---|---|---|---|---|
Region | Dependent variable | rRMSE | MSA | rME | Precision | Correlation |
Polar | |Grad_Ne@100km| | 0.47 | 135 | 0.00 | 0.37 | 0.36 |
|Grad_Ne@50km| | 0.49 | 136 | 0.00 | 0.37 | 0.35 | |
|Grad_Ne@20km| | 0.58 | 141 | 0.00 | 0.31 | 0.30 | |
IPIR_ix | 0.24 | 115 | 0.00 | 0.46 | 0.45 | |
Ne | 0.16 | 110 | 0.00 | 0.76 | 0.75 | |
Auroral | |Grad_Ne@100km| | 0.47 | 135 | 0.00 | 0.31 | 0.31 |
|Grad_Ne@50km| | 0.18 | 112 | 0.00 | 0.28 | 0.28 | |
|Grad_Ne@20km| | 0.98 | 168 | 0.01 | 0.26 | 0.24 | |
IPIR_ix | 0.24 | 115 | 0.00 | 0.37 | 0.36 | |
Ne | 0.15 | 110 | 0.00 | 0.76 | 0.75 | |
Mid-latitude | |Grad_Ne@100km| | 0.19 | 112 | 0.00 | 0.24 | 0.23 |
|Grad_Ne@50km| | 0.19 | 113 | 0.00 | 0.22 | 0.21 | |
|Grad_Ne@20km| | 0.98 | 158 | 0.00 | 0.20 | 0.20 | |
IPIR_ix | 0.30 | 126 | 0.00 | 0.33 | 0.34 | |
Ne | 0.16 | 111 | 0.00 | 0.73 | 0.68 | |
Equatorial version A | |Grad_Ne@100km| | 0.41 | 131 | 0.00 | 0.32 | 0.32 |
|Grad_Ne@50km| | 0.20 | 114 | 0.00 | 0.34 | 0.30 | |
|Grad_Ne@20km| | 1.15 | 169 | −0.01 | 0.46 | 0.26 | |
IPIR_ix | 0.32 | 122 | 0.00 | 0.27 | 0.26 | |
Ne | 0.15 | 111 | 0.00 | 0.55 | 0.55 | |
Equatorial version B | |Grad_Ne@100km| | 0.43 | 131 | −0.05 | 0.57 | 0.30 |
|Grad_Ne@50km| | 0.21 | 114 | −0.02 | 0.56 | 0.29 | |
|Grad_Ne@20km| | 1.16 | 170 | −0.12 | 0.61 | 0.28 | |
IPIR_ix | 0.34 | 123 | −0.04 | 0.48 | 0.19 | |
Ne | 0.16 | 111 | −0.03 | 0.83 | 0.53 |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.