Hi, The default value for lag.max in ACF implementation is 10*log10(N) There several publications recommending setting lag.max to: - N/4 (Box and Jenkins, 1970; Chatfield, 1975; Anderson, 1976; Pankratz, 1983; Davis, 1986; etc.) - sqrt(N)+10 (Cryer, 1986) - 20<=N<=40 (Brockwell and Davis) Why R uses 10*log10(N) as a default? Please, give me a reference to a book or article where the recommendation for using lag.max=10*log10(N) is proposed and explained. Thanks Rafal
I don't know why the default lag.max is 10*log10(N/m) for m series.
The "acf" help page includes the following:
Author(s):
Original: Paul Gilbert, Martyn Plummer. Extensive modifications
and univariate case of 'pacf' by B.D. Ripley.
I've copied these three contributors on this email. With luck, one
of them will enlighten us both.
Best Wishes,
spencer graves
Rafal Stankiewicz wrote:
> Hi,
>
> The default value for lag.max in ACF implementation is 10*log10(N)
>
> There several publications recommending setting lag.max to:
> - N/4 (Box and Jenkins, 1970; Chatfield, 1975; Anderson, 1976;
> Pankratz, 1983; Davis, 1986; etc.)
> - sqrt(N)+10 (Cryer, 1986)
> - 20<=N<=40 (Brockwell and Davis)
>
> Why R uses 10*log10(N) as a default?
>
> Please, give me a reference to a book or article where the
> recommendation for using lag.max=10*log10(N) is proposed and explained.
>
> Thanks
> Rafal
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
The default is taken from S-PLUS, so the reference is the S-PLUS manual. It is pretty similar to the recommendation of Brockwell & Davis. On Mon, 27 Mar 2006, Spencer Graves wrote:> I don't know why the default lag.max is 10*log10(N/m) for m series. > The "acf" help page includes the following: > > Author(s): > > Original: Paul Gilbert, Martyn Plummer. Extensive modifications > and univariate case of 'pacf' by B.D. Ripley. > > I've copied these three contributors on this email. With luck, one > of them will enlighten us both. > > Best Wishes, > spencer graves > > Rafal Stankiewicz wrote: > >> Hi, >> >> The default value for lag.max in ACF implementation is 10*log10(N) >> >> There several publications recommending setting lag.max to: >> - N/4 (Box and Jenkins, 1970; Chatfield, 1975; Anderson, 1976; Pankratz, >> 1983; Davis, 1986; etc.) >> - sqrt(N)+10 (Cryer, 1986) >> - 20<=N<=40 (Brockwell and Davis) >> >> Why R uses 10*log10(N) as a default? >> >> Please, give me a reference to a book or article where the recommendation >> for using lag.max=10*log10(N) is proposed and explained. >> >> Thanks >> Rafal-- 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