Skip to main content

Modeling Trends

Discussion Paper 22 | Published December 1, 1989

Download PDF

Author

Christopher A. Sims

Modeling Trends

Abstract

Models of low-frequency behavior of time series may have strongly conflicting substantive implications while fitting the data nearly equally well. We should develop methods which display the resulting uncertainty rather than adopt modeling conventions which hide it. One step toward this goal may be to consider “overparameterized” stationary ARMA models.