In this paper we investigate Turkish inflation forecast performance for nine alternative statistical models, particularly focusing on the effects of both linearity and stationarity. Moreover, out of sample forecasts obtained through the models having minimum root mean square errors are combined with using fixed and varying weights approaches. We conclude that the combination of forecasts of the nonstationary artificial neural network and the nonstation-ary vector autoregressive model have the best one-ahead forecast performance for Turkish inflation.
Inflation forecast VAR Artificial Neural Network Nonlinear Models
Diğer ID | JA69NC26TD |
---|---|
Bölüm | Makaleler |
Yazarlar | |
Yayımlanma Tarihi | 1 Aralık 2003 |
Yayımlandığı Sayı | Yıl 2003 Cilt: 5 Sayı: 3 |