Table 1 |
||||
|
Statistics on model performance |
||||
|
Male |
Female |
|||
|
|
||||
|
Sample |
Root mean squared error |
Correlation |
Root mean squared error |
Correlation |
|
|
||||
|
1000 |
1.18 |
0.85 |
1.03 |
0.78 |
|
|
||||
|
2000 |
1.14 |
0.86 |
0.94 |
0.83 |
|
|
||||
|
5000 |
0.96 |
0.91 |
0.86 |
0.86 |
|
|
||||
|
7000 |
0.95 |
0.91 |
0.79 |
0.89 |
|
|
||||
|
10000 |
0.79 |
0.94 |
0.79 |
0.88 |
|
|
||||
|
50000 |
0.49 |
0.98 |
0.55 |
0.95 |
|
|
||||
|
Sample indicates the size of the small county drawn from large counties, where sampling error of death rates is negligible. Root mean squared error and correlation are all relative to the (true) life expectancy of the large county. See Methods for further details. |
||||
|
Kulkarni et al. Population Health Metrics 2011 9:16 doi:10.1186/1478-7954-9-16 |
||||
