The Optimal Use of Provisional Data in Forecasting With Dynamic Models
提出一种基于状态空间和卡尔曼滤波的方法，在动态模型中高效利用临时数据（如工业产值初值），并以意大利数据为例，证明该方法优于传统预测方式。
Timely economic forecasts by means of dynamic models rely on updated time series, the last figure(s) of which are provisional and will be typically subjected to a number of revisions. A general approach to the efficient use of provisional observations in dynamic models is presented, based on the state-space methodology and the Kalman filter. Suitable adaptations are introduced, chiefly involving the measurement equations. Some applications are carried out for Italy, concerning (a) the monthly index of industrial production and (b) a small dynamic simultaneous-equation model of the aggregate economy. Kalman-filter estimates and predictions are compared with more traditional procedures.