regression with garch specification

Hi, it is not a technical question on matlab, but a doubt on garch model. Brefly: I have regressed the absolute return dispersion(CSAD) against a constant and the return, I didi it using Newey-west specification and garch model and teh last one gave better results, since the error term follows a garch model. Now I would like to see if the coefficient of the regression changes when we consider the rising and declining market separetely, my question is, may I do it using GARCH model or the faact that the sample is split create problem. In this case what do you think if, instead of splitting there sample in rising and declining I use an equation with dummies like this: CSAD=a+B*R+B2*D*R, where D is one when the return is negative, the B should represent teh coeffcient of the up market and B+B2 the one of the down market, right. And this should not create any problem to GARCh.

Thanks in advance Gaia

[Oleg Komarov]

Teh function accepts regressors in the mean equation if this is what you're asking.