Page 1 of 1

Overflow error

Posted: Thu Dec 08, 2011 1:07 pm
by zngtian
I tried to run simultaneous equation system and forecast. In the forecast I encountered an overflow error message, saying that "Overflow in DO_ECIO.SOLVE(O=G)". ECIO is the name of the model. I tried to scale down some variables but it didn't work. Could someone please help me out or give a clue on how to resolve this. Thanks,

Zheng Tian

Overflow error

Posted: Thu Dec 08, 2011 1:21 pm
by EViews Gareth
Do you have logs in the specification? It can be quite easy to get an overflow when dealing with logs.

Re: Overflow error

Posted: Thu Dec 08, 2011 1:41 pm
by zngtian
Yes, I have some variables that need to be transformed to log. The solution message of the ECIO model is like that

"Solve begin 15:01:29
2015Q4 NA generated for XIO15 - Overflow"

But XIO15 is not logarithmic. The forecast before 2015Q4 is good but in 2015Q4 all variables seem to stop there and accumulate to very big numbers.

Overflow error

Posted: Thu Dec 08, 2011 1:49 pm
by EViews Gareth
It sounds as though your system is explosive and is simply getting too big.

Re: Overflow error

Posted: Thu Dec 08, 2011 2:15 pm
by zngtian
But it doesn't seem to be the out-of-memory problem. Is there any way to solve it?

Re: Overflow error

Posted: Thu Dec 08, 2011 2:19 pm
by EViews Gareth
What do you mean by solve it? Your system is explosive. That's a valid result. It may not be the result you want, but there's no way to "solve" it without re-specifying your system.

Re: Overflow error

Posted: Thu Dec 08, 2011 2:36 pm
by zngtian
OK. I'll work on change some equations in the system. Thanks for help

Re: Overflow error

Posted: Tue Aug 13, 2019 2:26 pm
by geoanstats
Gareth- what does it mean that a system is explosive? I don't understand the terminology.

Thanks

Overflow error

Posted: Tue Aug 13, 2019 3:30 pm
by EViews Gareth
Generates number that are huge.

Imagine you have a model that says:
Y=10*y(-1)^3

With y_0=1000 and you forecast it for 100 periods - you’ll find you quickly get to a point where the numbers are infeasibly large.