Message-ID: <518A33D3.1050601@braverock.com>
Date: 2013-05-08T11:15:31Z
From: Brian G. Peterson
Subject: rugarch VaR calculation "manually"
In-Reply-To: <CADih64QudPg=Ur2kz9TsUcMQBD1jqfNCzLqPfv_Gc1A-r6RTug@mail.gmail.com>
On 05/08/2013 06:06 AM, Neuman Co wrote:
> I want to use all observations for my
> fitting and then do one step ahead in sample predictions.
There is no such thing as a one step ahead in-sample *prediction*.
Please stop and think a moment about how absurd that is before you waste
more of everyone's time.
At each time t, I may make a prediction for time t+1
ugarchroll will give you a set of rolling one or n step ahead
*predictions*, as Alexios has also already told you.
If you want the in-sample *estimate* from the fitted model values (it is
not a prediction), then you want the 'fitted' and 'sigma' values, as
Alexios has told you at least twice now.
...and I still don't see you signing your name.
Brian
--
Brian G. Peterson
http://braverock.com/brian/
Ph: 773-459-4973
IM: bgpbraverock