[R] 0-only-rows in correspondence analysis

[Stephane DRAY]

CA is based on relative composition and not on absolute value. So, empty 
site do not give any information. In a computational point of view, empty 
row or column do not allow to compute row / col weights. that is why most 
function do not work with empty row or column. dudi.coa (ade4 package) 
allows empty rows or columns. They are simply put at the centroid (and so 
have no weight in the analysis). CA do not take into acccount the 
information given by empty site !!! If you want that the information given 
by empty site are taken into account, you should prefer an other analysis 
(e.g. pca).

Sincerely,


At 10:09 22/10/2003, you wrote:
>Hello,
>
>I am looking for a workaround of "species empty" plots in a correspondence
>analysis.
>
>I tried to do a community structure analysis with a ca via decorana(), ca()
>and CAIV(), but none of them allow 0 only-rows.
>
>I have species (abs./pres) in columns and sites in rows
>
>         sp1 sp1 sp3 ...
>plot1   0    1     1
>plot2   0    0    0
>plot3   1    0     1
>.....
>
>but on some plots no species could be found, therefore only 0 appear. I know
>that this isn't interesting for a community analysis but I am interested in
>the contributions of variables to the factors for a further analysis with
>env. variables where it is interesting that there are no species found.
>
>i thought about adding a "dummy" species with only presence on all sites, 
>that
>would solve the 0-only-row problem but I am not sure if that would bias the
>outcome of a ca seriously.
>
>Or is there a subcommand to allow 0 rows in a ca which I haven't seen?
>
>thnaks Martin
>
>______________________________________________
>R-help at stat.math.ethz.ch mailing list
>https://www.stat.math.ethz.ch/mailman/listinfo/r-help