BEKK estimation on residuals obtained from ARMA fit

[bonjourbc9]

Hi Folks,
I have a quick question on the diagonal BEKK estimation in EVIEWS. Suppose I have 2 sets of daily log returns ( S&P500 , FTSE100) I call it Rs&p and Rftse and I fitted a ARMA(1,0) model as the conditional mean for both returns;

Rs&p(t) = c1 + c2Rs&p(t-1) + e1t
Rftse(t)=c3 + c4Rftse(t-1) +e2t


Next I extract the residuals from these conditional mean equation and I want to perform a diagonal BEKK to model the dependence between these pair of residuals (e1t,e2t).

If I use the system estimation in EVIEWS and choose Diagonal BEKK model, will EVIEWS assume a conditional mean again for my residuals ? Eg. will it assume a conditional mean equation of the form e1t = c1 + u1t and e2t=c2 + u2t ?? How to I ensure that EVIEWS treat my data as follows ; e1t = u1t and e2t=u2t.

Appreciate your input. Thanks.

[bonjourbc9]

Hi Folks,
I have a quick question on the diagonal BEKK estimation in EVIEWS. Suppose I have 2 sets of daily log returns ( S&P500 , FTSE100) I call it Rs&p and Rftse and I fitted a ARMA(1,0) model as the conditional mean for both returns;

Rs&p(t) = c1 + c2Rs&p(t-1) + e1t
Rftse(t)=c3 + c4Rftse(t-1) +e2t


Next I extract the residuals from these conditional mean equation and I want to perform a diagonal BEKK to model the dependence between these pair of residuals (e1t,e2t).

If I use the system estimation in EVIEWS and choose Diagonal BEKK model, will EVIEWS assume a conditional mean again for my residuals ? Eg. will it assume a conditional mean equation of the form e1t = c1 + u1t and e2t=c2 + u2t ?? How to I ensure that EVIEWS treat my data as follows ; e1t = u1t and e2t=u2t.

Appreciate your input. Thanks.

Hi Folks, while waiting for someone to reply, I figured the following. Suppose I load the set of errors e1t and e2t into EVIEWS as series01 and series02. When I hightlight these 2 series and select Open -- As system and then under the spec , I define the equation as

@STACKINST

@INST

SERIES02

SERIES01

This should prompt EVIEWS to view series01 as error terms and series02 as error terms right? So if I proceed and estimate Diagonal BEKK, I would infact have performed a DIAGONAL BEKK estimaton on the error terms e1t and e2t ? Please advise.

[trubador]

Do not try to drop the constant terms in conditional variance equations. They are important components in terms of ensuring the nonnegativity...

[bonjourbc9]

Thanks. I did otherwise. Instead of fitting the diagonal BEKK to the residuals obtained from each conditional AR mean equation, I define the system first hand in EVIEWS as

@STACKINST

@INST

SERIES02 = C(1) + c(2)*SERIES02(-1)

SERIES01= C(3) + C(4)*SERIES01(-1)

where both series are now the daily log returns of SP500 and FTSE 100 instead, afterwhich I estimate system by fitting the diagonal BEKK.