ARMA to Excel

[jgutierr]

Hi,
We are working on some ARMA models and then replicating the results for yhat in Excel. In order to do so I have followed the advice given in an EViews forum (http://forums.eviews.com/viewtopic.php?f=7&t=465).

They say that in order to replicate the results of an ARMA model in Excel, one have to take into account that the actual estimated model is:

y_t = alpha + beta*x_t + u_t
u_t = rho*u_t-1 + epsilon_t.

After replacing it is possible to reach the following expression:

y_t = alpha + rho*(y_t-1 - alpha – beta*x_t-1) + beta*x_t + epsilon_t.

I’m a little confused on how to replace the terms when I have the following model:

y_t = alpha + theta*y_t-5 + beta*x_t-4 + u_t

Should I define u_t as rho*u_t-5 + epsilon_t, so as to have:

y_t = alpha + beta*x_t-4 + u_t,
u_t = rho*u_t-5 + epsilon_t, and
y_t = alpha + rho*(y_t-5 - alpha – beta*x_t-9) + beta*x_t-4 + epsilon_t.

I've tried it but didn't reach the estimate of Eviews.
Am I doing something wrong?

I appreciate your kind help.
Regards,
JGR