Table 2 Out-of-sample results. Comparison of models in Table 1 with respect to out-of-sample-predictions. Mean square error is estimated via randomly sampling 80% of the data to estimate the model, using the remaining 20% to test the model, and repeating 100 times. SD = standard deviation. The best out-of-sample prediction statistics are reported in bold
From: Public debate in the media matters: evidence from the European refugee crisis
| Â | Out-of-sample prediction measures | |||
|---|---|---|---|---|
| Â | Mean absolute error | Mean square error | ||
|  | | Predicted% − True% | | (SD) | (Predicted% − True%)2 | (SD) |
Media models from Table 1 |  |  |  |  |
Model (1) | 3.08% | (0.05) | 17.2 | (0.6) |
Model (2) | 3.07% | (0.1) | 17.1 | (0.6) |
Control models from Table 1 |  |  |  |  |
Model (3) | 35.4% | (0.3) | 1507.4 | (20.0) |
Model (4) | 35.4% | (0.3) | 1504.3 | (17.7) |
Full models from Table 1 |  |  |  |  |
Model (5) | 35.4% | (0.3) | 1504.1 | (19.7) |
Model (6) | 35.4% | (0.3) | 1504.5 | (17.0) |