R help - Mar 2004 - calculate length of gradient ?

Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Virus-Scanned: by amavisd-new
X-Spam-Checker-Version: SpamAssassin 2.63 (2004-01-11) on hypatia.math.ethz.ch
X-Spam-Level: ****
X-Spam-Status: No, hits=4.2 required=5.0
tests=MSGID_FROM_MTA_HEADER,RCVD_IN_BL_SPAMCOP_NET autolearn=no version=2.63

Dear r-help list,

my question is about ordination technics:
To decide using a linear model response or a unimodal model response,
in Canoco (software for ordination technics)
you can calculate a so-called "length of gradient".

(from Canoco Help file:)
> The length of gradient is a measure of how unimodal the species
> responses are along an ordination axis. It is the range of the
> sample scores divided by the average within-species standard
> deviation along the axis. The gradient length is expressed in
> standard deviation units of species turnover (SD).
The function  cca in package vegan and other functions give many
outputs, but I didn't find anything about "length of gradient". Is
there a possibility to calculate this in R? If yes how can I
calculate this?

Many thanks in advance, Andreas
_______________________________________________________________________
... and the winner is... WEB.DE FreeMail! - Deutschlands beste E-Mail

On 29 Mar 2004, at 16:54, Andreas Plank wrote:on=2.63
> my question is about ordination technics:
> To decide using a linear model response or a unimodal model response,
> in Canoco (software for ordination technics)
> you can calculate a so-called "length of gradient".
>
> (from Canoco Help file:)
>> The length of gradient is a measure of how unimodal the species
>> responses are along an ordination axis. It is the range of the
>> sample scores divided by the average within-species standard
>> deviation along the axis. The gradient length is expressed in
>> standard deviation units of species turnover (SD).
>
> The function  cca in package vegan and other functions give many
> outputs, but I didn't find anything about "length of
gradient". Is
> there a possibility to calculate this in R? If yes how can I
> calculate this?
You cannot get this scaling in the cca function. However, function 
decorana() gives you "axis lengths". Cajo ter Braak suggested a linear
approximation to Mark Hill's "sd scaling" that would be easy to 
implement in cca (actually, in summary.cca(), since no scaling is done 
until your request one), but I haven't bothered to do that. By the way, 
I disagree with the recommendation: CA is in general better than PCA, 
even with "short gradients" so  you don't need axis lengths.

Please read the posting guide (about messages that should be sent to 
the list vs. to the package authors).

cheers, jari oksanen
--
Jari Oksanen, Oulu, Finland