GARCH constraints not met?

[finnecon]

I am trying to use GARCH(1,1) to estimate the parameters for the long-term volatility.

I am following John Hull's method of calculating the long-term volatility.
The long-term volatility is (Coefficient of C) / [1 - (Coefficient for RESID(-1)^2) - (Coefficient for GARCH(-1)]
subsituting w for the Coefficient of C
a for the Coefficient for RESID(-1)^2
b for the Coefficient for GARCH(-1)
You get, Long-Term Variance = w / (1-a-b).

I imported the return data into eviews where return(t) = price(t) / price(t-1) and ran the ARCH method.
Under mean equation, i put, "return c" and leaving all else as default, I pressed OK.

Running this under different time intervals, w, a, and b are sometimes negative.
On some instances, the sum of a and b are greater than 1, which causes the long-term volatility to become "mean fleeting".

How do I adjust the estimation parameters so that the following two constraints are always met?
Constraint 1: a > 0, b > 0, w > 0
Constraint 2: a + b < 1

If anyone sees me doing anything wrong here, please advise. Thanks in advance.

[EViews Gareth]

EViews doesn't let you impose constraints on coefficients. However, if you do a forum search you'll see there are a couple of tricks you might be able to use.

[startz]

EViews doesn't let you impose constraints on coefficients. However, if you do a forum search you'll see there are a couple of tricks you might be able to use.

Is it possible to specify explicit coefficients for the GARCH terms?

[EViews Gareth]

Ah, good point, it is not, and I apologise for not reading the original post thoroughly enough!