On comparing multi-horizon forecasts

被引:5
作者
Capistran, Carlos [1 ]
机构
[1] Banco Mexico, Mexico City 06059, DF, Mexico
关键词
evaluating forecasts; forecasting practice; multi-horizon forecasting;
D O I
10.1016/j.econlet.2006.04.010
中图分类号
F [经济];
学科分类号
02 ;
摘要
When comparing multi-horizon forecasts, the common method of taking the square of the forecast errors for each horizon and then averaging over the horizons performs worse, in terms of the user's expected loss, than applying a Diebold-Mariano test using a multivariate loss function. (c) 2006 Elsevier B.V. All rights reserved.
引用
收藏
页码:176 / 181
页数:6
相关论文
共 5 条
[1]  
BANERJEE A, 2003, UNPUB EUROPEAN FOREC
[2]  
CECCHETTI SG, 2000, CURRENT ISSUES EC FI, V6, P1
[3]  
Corradi V, 2004, INT J FORECASTING, V20, P185, DOI [10.1016/j.ijforecast.2003.09.008, 10.1016/j.ijforecast.2004.09.008]
[4]  
DEN HAAN W.J., 1997, Robust Inference, V15, P299
[5]   COMPARING PREDICTIVE ACCURACY [J].
DIEBOLD, FX ;
MARIANO, RS .
JOURNAL OF BUSINESS & ECONOMIC STATISTICS, 1995, 13 (03) :253-263