[R] Non-metric multi-dimensional scaling with "NA"s / "0"s

[Sarah Goslee]

What do you mean by "ignored"? Do you want to drop those sites
from the ordination? If so, you need to subset your data beforehand.

Or if you mean that you have presence data, and not really absences
(common in ecology), and want to ignore joint absences, then you
need to select a dissimilarity metric that does that. Bray-Curtis maybe?

Sarah

On Mon, Feb 15, 2010 at 4:38 PM, kellys17 <seankelly999 at hotmail.com> wrote:
>
> Hi,
>
>
>     I am tryng to use Non-metric multi-dimensional (MDS) to illustrate the
> difference in mixed-species flock composition between two species of birds.
> It is part of a thesis I am currently working on.
>
>  My data consists of 18 columns and 62 rows, with numbers ranging from 0 to
> 7. I have carried out MDS of the data various ways but when "0" is kept in
> the data set uses "0" as a measure when I want it to be ignored. I have
> tried to counter this by replacing all "0"s with "NA" but to no avail. I am
> struggling to find a function which will allow me to carry out MDS without
> taking into account "NA"s or "0"s, essentially carry out MDS with "missing
> values".
>
>  If anyone could advise me on my problem it would be much appreciated.
>
>
>  Thank you,
>
>


-- 
Sarah Goslee
http://www.functionaldiversity.org