The ECMWF temperature and precipitation ensemble reforecasts are evaluated for biases in the mean, spread and forecast probabilities, and how these biases propagate to streamflow ensemble forecasts.
These include metrics that verify the forecast probabilities and metrics that verify the ensemble mean forecast.
The repeaters occurred from 1993 until the forecast time were used to calculate the forecast probability.
The forecast probability for the patients in the preliminary study was calculated from their current sick-listing at baseline.
Table 5 shows the mean forecast errors for both the categorical and continuous variable models.
For the analysis we consider seasonal mean forecast anomalies for December to February (DJF) from forecasts started on 1 November each year.
The fit of each multivariate model was evaluated with the Hosmer-Lemeshow χ2 goodness-of-fit test, and its discriminatory power by means of the area below the ROC curve, taking into account the forecast probabilities and the variable status (high score/low score), with a cut-off point at p = 0.5.
Critically, these forecast probabilities were just that -- forecast probabilities -- not model-based, poll-average probabilities of an Obama lead in the polls.
The probabilities we were displaying on the maps on the Pollster dashboard during the campaign were the more confident, "probability of Obama lead" quantities, not the forecast probabilities shown here.
Only during 2000 did forecast probabilities noticeably exceed those of climatology.
