On Forecasting certain Misspecified Models
O. D. Anderson · Journal of Information and Optimization Sciences · 1981
It is believed that “similar” nearly nonstationary models and nonstationary approximations to them can frequently be distinguished—but the question is tben whether implementing the discrimination leads to better forecasts. In this paper we consider one such pair of “similar” models anddemonstrate that the procedure is worthwhile, although the gains may not be spectaculltr. This provides another witness in the ca~e against differencing series unnecessarily.