Non-negativity constraints GARCH(1,1)

[hendrix]

I'm running a GARCH(1,1) model on intraday data with a bunch of dummy variables in the variance equation (time-of-day, day, news etc.). The problem is that I get negative coeffisient for some of the dummy's. Is it possible constrain the paramters to be non-negative?

EViews 7.1

[EViews Gareth]

http://forums.eviews.com/viewtopic.php?f=4&t=48&p=103

[hendrix]

Thank you,

But I can't get it to work when using a GARCH model, probably because I'm not familiar with estimation using the command window. Anyways my mean equation is:

Code: Select all

return c return(-1)

and the variance regressors which I need to constraint is:

Code: Select all

group1 group2 group3

[EViews Gareth]

Sorry, didn't realise you wanted to constrain the variance regressors. That can't be done.

[hendrix]

Ok.

I'm not too experienced with GARCH-modelling. My lagged squared error and lagged variance term is non-negative, is it acceptable that some of the dummys coeffisients is negative?