Bayesian Composite Qualitative Forecasting: Hog Prices Again
提出一种新的组合定性预测方法，利用辅助logit模型评估各定性预测的正确概率，再基于这些概率的归一化值组合预测。以生猪价格方向变化预测为例，证明该方法无需知道单个预测的生成方式即可形成更优的组合预测。
Abstract A new method for forming composite qualitative forecasts is presented. A set of qualitative forecasts is evaluated using auxiliary logit models to predict the probability of each forecast's correctness. Individual model forecasts are then combined on the basis of normalized values of these probabilities. This method is demonstrated with three sets of forecasts on the direction of change in hog prices (up or down). The application shows that without any information on the manner in which the individual forecasts are generated this method can form a composite forecast that is superior according to a variety of metrics for evaluating qualitative forecasts.