Out-Of-Sample-Forecasting SETAR-model

Potti_X · Postby **Potti_X** » Tue Aug 21, 2012 4:28 am

hi there,
im trying to do some out-of-sample forecasts for my little SETAR-like model.


series y = (gbpusd)
series growth = dgbp

scalar th = -0.04
series ind = (growth(-1)<th)

equation test.ls y c ind*y(-1) (1-ind)*y(-1)

the in-sample forecasts works perfectly fine, i already expanded the workfile but when i run a dynamic out-of-sample forecasts it only shows the first out of sample observation forecast (e.g. we are in 2012M08 so i forecast to 2013M08, it only shows me the forecast for 2012M09 and thats it). Is there any way to fix this within the Forecast Interface or should i use monte carlo/ bootstrap methods?

thx in advance & best regards

EViews Gareth · Postby **EViews Gareth** » Tue Aug 21, 2012 7:41 am

Does ind exist for your forecast period?

Potti_X · Postby **Potti_X** » Tue Aug 21, 2012 8:10 am

no, thats what i also suspect why its not working....

the problem is that ind depends on y, so i would need the forecasts for y in order to get data for ind...
do u have a small trick or roundabout for this problem? using AR(q) forecasts and use them to calculate ind seems a bit quick & dirty

thx for your help

regards

EViews Gareth · Postby **EViews Gareth** » Tue Aug 21, 2012 8:19 am

If IND depends upon Y, you'll need a model object to do the forecast, with an equation for IND.

Potti_X · Postby **Potti_X** » Wed Aug 22, 2012 1:37 am

thx gareth, works perfectly fine now. is there any possibility to calculate RMSE statistics like in the standard forecast window or do i have to calculate those by hand?

regards

EViews Gareth · Postby **EViews Gareth** » Wed Aug 22, 2012 6:37 am

There is an @rmse function for calculating the RMSE between two series.

EViews.com