# Finding ARMA coefficients

Started by April 7, 2006
```Hello! I have a problem to understand ARMA filter dsign.
I would be thankful to tell me in detail how can find coefficients of the
ARMA filter for any time-series data.

Thank you.

```
```"csksms" <kscsms123@yahoo.com> wrote in message
news:BM2dnU2bHJgLGKvZnZ2dnUVZ_tOdnZ2d@giganews.com...
> Hello! I have a problem to understand ARMA filter dsign.
> I would be thankful to tell me in detail how can find coefficients of the
> ARMA filter for any time-series data.
>
> Thank you.
>
>

Depends - do you know the white noise input? If so use recursive least
squares (RLS) otherwise use extended RLS (recursive maximum likelihood).

Tam

```
```csksms wrote:

> Hello! I have a problem to understand ARMA filter dsign.
> I would be thankful to tell me in detail how can find coefficients of the
> ARMA filter for any time-series data.

This is a newsgroup, not a library. ARMA modeling is a topic which is
usually covered in a whole book (or at least several chapters thereof)
and you can't expect anybody to sit down and recite the whole material
here. Get yourself a text book, read up, and come back with more
specific questions.

```
```>csksms wrote:
>
>> Hello! I have a problem to understand ARMA filter dsign.
>> I would be thankful to tell me in detail how can find coefficients of
the
>> ARMA filter for any time-series data.
>

hello

thats right, you need to read a chapter on adaptive filtering.

assuming you are talking about system identification using an arma filter,
try the steiglitz mcbride method. it requires few iterations, since the
work required for each iteration is quite a bit. however it is pretty
unstable for a signal of very high order (IMO 10 or above)

there is a paper from paulo diniz published in 1994 (dont have its link)
which explains the method pretty well. you could google it. good luck :-)
```