Reply by Syed Ahmar Shah July 24, 20132013-07-24
With regards to what is an 'optimal' value is debatable. There are a number
of methods suggested in the literature for this. For example, there is the
Akaike Information criteria (AIC), then there is identifying the lowest
model order before the error terms approach an asymptotic behaviour and a
number of other methods. My advice: Choose the lowest model order that
seems to do your job..
You can also choose a subset of data with known labels/answers and then
determine the 'optimal' model order.
hope this helps.
Best Wishes,
Ahmar
On Sun, Jun 30, 2013 at 7:56 PM, wrote:

> **
> Deal all;
> I have used the Yule-Walker method (pyulear) to get a PSD (i.e. [Pxx,f] > pyulear(x,p,nfft,fs) ). p is the order of an AR model for the sign. How can
> I determine the optimal value of p?
>
>

--
Syed Ahmar Shah
Reply by sdhs...@gmail.com July 2, 20132013-07-02
Deal all;
I have used the Yule-Walker method (pyulear) to get a PSD (i.e. [Pxx,f] = pyulear(x,p,nfft,fs) ). p is the order of an AR model for the sign. How can I determine the optimal value of p?