[R] What's wrong with ar for my data?

[Prof Brian Ripley]

You are apparently fitting a series for which the selected order is zero,
and ar.burg is not designed to cope with that (and would in any case tell 
you nothing useful).  The default method does cope, from your output.

Why are you fitting an AR model to a series with apparently no 
correlation?

On 19 Jun 2003, zhu wang wrote:

> Dear helpers,
> 
> When I use ar to fit the data with length 180, I have the following
> error:
> 
> ar(x,method="burg")
> Error in acf(x, type = "covariance",lag.max=order,plot=FALSE):
>          lag.max must be at least 1
> 
> If I use 
> 
> ar(x), then I have
> 
> Call (x=x)

[I very much doubt that is what you get.]

> 
> 
> order selected 0 sigma^2 estimated as 5374
> 
> Obviously I missed some points for using ar. 
> This is R 1.7.0 under Redhat Linux 9.0

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595