[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Problem with ARIMA-ARCH model

From   Robert A Yaffee <[email protected]>
To   [email protected]
Subject   Re: st: Problem with ARIMA-ARCH model
Date   Mon, 18 Aug 2008 13:07:38 -0400

   Sometimes when the variance model is more properly specified, the mean model
parameter estimates emerge as similar although not identical.   If the response surface 
of the nonlinear model has ridges, meadows, or mutliple modes,  you might wind up in 
a local optima, which could lead to different results.  This is a common problems
with nonlinear models such as the GARCH type.  You might try other specifications of your
GARCH model.  I hope that helps.
    - Regards,
              Bob Yaffee


Robert A. Yaffee, Ph.D.
Research Professor
Silver School of Social Work
New York University



----- Original Message -----
From: vasja sivec <[email protected]>
Date: Monday, August 18, 2008 12:39 pm
Subject: Re: st: Problem with ARIMA-ARCH model
To: [email protected]
Cc: [email protected]

> Dear Mr. Rober A. Yaffee, Mr. Brian Poi and Jao Lima,
> Thank you for your help, but sadly this still doesn`t improve the results.
> Mr. Yaffee and Mr. Poe suggested using the conditional option with
> garch command, because it could be that I was using unconditional ML
> with ARIMA mean model and then conditional ML when ARIMA was jointly
> estimated with ARCH, or vice verse. I tried running the following
> commands (hope they are right):
> . arch dlndax, arima(1,0,1) noconstant arch(1) nocondition nolog
> . arch dlndax, arima(1,0,1) noconstant arch(1) condition nolog
> The results between just ARIMA model and ARIMA model calculated
> jointly with an ARCH model were still different.
> The other suggestion was that the parameter space may be different
> with the ARCH model, so I should try random starting values or
> different starting values. But I really don`t know how to do that.
> Could it be that the commands are somehow misspecified?
> And thank for your help and suggestions!
> Kind regards,
> Vasja Sivec
> *
> *   For searches and help try:
> *
> *
> *
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index