<x-charset ISO-8859-1>In fact, the procedure described in Resh et al. (1994) was rather empirical
since we had not yet an achieved solution for the question of relating
environmental data to species traits. This question has many solution and is
very rich both on the biological (Ribera et al., Effect of land disturbance
and stress on species traits: a 3-table analysis of ground beetle
assemblages. Ecology in press) and statistical side (Dolédec et al.,
Matching species traits to environmental variables: a new 3-table ordination
method, EES, 3, 143-166).
To stay on the multivariate side as suggested by your experimental context,
I would not follow the Resh et al. (1994) procedure since we made some
substantial progress since then. A "straightforward solution" to your
question is to run an RLQ analysis on your three tables. In the present
state of the program this requires to analyse the abundance table by COA
(rather than row weighted COA which we used to be homogeneous with FCA). If
you consider it as too complex, you may compute a trait-by-site matrix (that
contains the proportion of trait modalities in sites). Note that the
analysis of such a matrix makes some kind of compromise among traits that
may obscure trends. A separate examination of traits may be required.
You may then match the trait-by-site with you environmental data. Note that
there is no reason to follow the procedure of Cellot et al. (1994) with your
own environmental data. The data that Cellot et al. (1994) had to manage we
very heterogeneous taken by various people on the long term and thus
required a procedure that reduced the measurement errors.
As you can see many things remains to do and it is not really possible to
give all the possibilities on this media. However many advises are also
available in the docs.
All the very best,
UCB Lyon 1
CNRS - Ecologie des Hydrosystèmes Fluviaux
43 Bd du 11 novembre 1918
phone: +33 (0) 472431363
fax: +33 (0) 472431141
>De : Byungok.Ahn@uni-essen.de
>À : email@example.com
>Objet : matching the results of coinertia-analysis with the scores of the
temporal and spatial variability
>Date : Lun 22 mai 2000 11:39
> I have been trying to test predictions of the river habitat templet
> concept based on 27 stream sites, 18 quantitative variables of temporal
> variability, 7 quantitative variables of spatial variability for each
> stream, 86 species in 6 scales (0 to 5), and 80 fuzzy coded species
> modalities in 22 categorical variables.
> Following the procedures in Resh et al. (1994) four tables were
> separately processed as follows:
> 1. A species traits table (86 rows, 80 columns) was processed by Fuzzy
> 2. An abundance (habitat utilization) table (86 rows, 27 columns) was
> processed by row-weighted COA.
> 3. Coinertia-analysis using the iita.file after matching two tables.
> 4. Non-centred PCA using two tables of temporal and spatial variability
> separately after transformations of all variables described by Cellot et
> al. (1994).
> My problem: I would like to examine and represent positions of the
> stream sites, species, and species modalities in the temporal and
> spatial habitat templet graphically, but I don=B4t know how to match the
> results of coinertia-analysis with the scores of the temporal and
> spatial variability. I think that a further coinertia-analysis may be a
> possible solution for it. But how can I prepare for it?
> Moreover I cannot find a file which may contain correlation ratios
> between each categorical variable and each factorial score of the
> Any help would be appreciated.
> Byungok Ahn
> Istitut fur Ekologie
> Abt. Hydrobiologie
> FB 09 Univ-GH-Essen
> Universit=E4tsstr. 5
> 45117 Essen
> e-mail: firstname.lastname@example.org
> Tel.: 0201/183-3201
> Fax: 0201/183-2529
This archive was generated by hypermail 2b30 : Mon Feb 12 2001 - 09:24:55 MET