Forecasting based on an ECM

Questions and discussions on Time Series Analysis
challenges
Posts: 11
Joined: Thu Apr 14, 2011 12:29 pm

Forecasting based on an ECM

Unread post by challenges »

Dear all,

I have a problem with forecasting results on the basis of an ECM. When I make dynamic forecasting on the basis of an ECM with Rats 6.2, I obtain graphs with observed changes, as well as forecasted changes of my endogenous variable for the period of forecasting. However, several articles provide graphs with the forecasted level as well as the observed value of the endogenous variable on the basis of an ECM. Could you please explain me the way I can obtain the forecasted level of the endogenous variable on the basis of an ECM, as it is not clear to me ?
Thank you in advance.
TomDoan
Posts: 7814
Joined: Wed Nov 01, 2006 4:36 pm

Re: Forecasting based on an ECM

Unread post by TomDoan »

If you do your VECM with SYSTEM with an ECT command, your forecasts will be for the levels, rather than differences. In you're estimating the VECM in differenced form, then the levels from the forecasts would be done something like:

set fy fstart fend = fdy+%if(t==fstart,y(fstart-1),fy{1})

where y is the original series, fdy is the forecast of the difference, fstart and fend gives the range of forecasts. fy is the forecasts of y.
Post Reply