Sign in
Increase in mean square forecast error when omitting a needed covariate
Journal article   Peer reviewed

Increase in mean square forecast error when omitting a needed covariate

Johannes Ledolter
International journal of forecasting, Vol.23(1), pp.147-152
01/2007
DOI: 10.1016/j.ijforecast.2006.10.001

View Online

Abstract

Mean square errors of ex-post and ex-ante forecasts from transfer function (regression) models are compared with mean square forecast errors of univariate time series models that ignore the covariate. We show that forecasts from the univariate ARMA models are never better, and are usually worse, than the forecasts from the transfer function model.
ARMA model Ex-ante forecast Ex-post forecast Transfer function model

Details

Metrics