On the integration of multi-step prediction and model selection for stationary time series
J.T.C. Kool, Arnold H. Q. M. Merkies · RePEc: Research Papers in Economics · 1986
An important aim of modelling economie time series is the generation of predictions.However a model, which is suitable for describing the data within a sample, may be suboptimal for the generation of out sample forecasts, especially for increasing time horizons.The objective of this paper is to analyze a model selection procedure, which is also based on the predictive performance.In first instance the time horizon of prediction is formulated and then the appropriate model is selected, so different time horizons of prediction may lead to different models.After a suitable model has been chosen, we also examine the method of estimation.The properties of the various estimators and the model selection procedure are investigated by simulation.