Page 1 of 1

Forecast evaluation criteria

Posted: Sat Jul 04, 2009 4:10 am
by Thomas
Hello all,

I am in the process of writing an assignment comparing the forecasting abilities of ppp, uip, dornbusch frankel with and without share prices in first-difference and error correction form for exchange rates. I am planning to look at the in-sample and out of sample forecasting abilities.

Does Eviews have the insample metrics :

Hit ratio, Kuipers Score, Pesaran-Timmerman statistic??

Does Eviews have the out of sample metrics :

(1) MSE ratio (the ratio between the mean squared error of the structural models and a driftless random walk is used) and the Diebold Mariano Test
(2) The Direction of change criterion (proportion of forecasts that correctly predict the direction of exchange rates) and
(3) the consistency criterion (that the forecast and actual realization comove one-to-one in the long run)??

And how can i find out the actual cointegrating vector in eviews in relevant tests?
If i cannot get these criteria from eviews...does anybody know whether Splus ( finmetrics included) provides these?

any help would be nice!!!!
Thanks

Re: Forecast evaluation criteria

Posted: Mon Mar 04, 2013 2:30 am
by sim
I am facing the same problem. Can somebody help?

Kind regards

Re: Forecast evaluation criteria

Posted: Mon Mar 04, 2013 6:00 am
by EViews Gareth
EViews has none of those built in, but it may be that they are easy to calculate.

Re: Forecast evaluation criteria

Posted: Mon Mar 04, 2013 7:36 am
by sim
How do you calculate the Direction of Change criterion?

The Diebold Mariano Test is explained in another thread, and hopefully I'll understand it, but I didn't find anything about the construction of the Direction of Change criterion.

Kind regards

Re: Forecast evaluation criteria

Posted: Mon Mar 04, 2013 8:04 am
by trubador
Gareth is right.
If I am not misunderstanding the definition, direction of change can be calculated easily as follows:

Code: Select all

'Suppose the name of dependent variable is y and its forecasted values are already stored in y_f
series match =  1-((d(y)>0 and d(y_f)<0) or (d(y)<0 and d(y_f)>0))
scalar doc = @sum(match)/match.@obs

Re: Forecast evaluation criteria

Posted: Thu Sep 07, 2017 1:05 pm
by robertvjanssen
Hi all,

I was wondering if the pesaran-timmermann test is built in the latest version of Eviews (10)?

I performed the pesaran-timmermann test following the equations stated above, but I'm looking for the any statistics which go with that, in order to determine any significance.

Could you help me out?

Thanks up front!

Robert

Re: Forecast evaluation criteria

Posted: Thu Sep 07, 2017 1:06 pm
by EViews Gareth
It is not.