>I have performed a Fuzzy Correspondence Analysis on a macrophyte species by
>species trait matrix. I wish to create diagrams similar to those in figure
>1, b of, Bornette et al
> (1994). The diagrams consist of species linked to their individual
>modalities. There is one diagram for each trait. I have tried doing this in
>Scatters and Scatterclass using the star option with:
> XY coordinate file = file.flli
>Categories file (*.cat) = fileFModa.cat
>The category file used, is the problem I think. It seems to be incorrect
>for this type of diagram. Any suggestions?
This question was already addressed by eg. Francisco Leonardo Tejerina
Garro. I hereunder give a short translation (as requested by Kyrre) of what
The paper of Bornette et al. used the version 3.7 of ADE that included 4
graphic modules for drawing fuzzy coded variables (Fuzzytable,
FuzzyRowscore, FuzzyScore1 and 2).
The Fuzzy Table option is available in the new version Tables: Fuzzy
Variables (see topic documentation 81).
The FuzzyRowScore module positioned trait modalities (columns) with their
scores of variance equal to 1. A species (row) was positioned at the
average of trait modalities of that species (row). By contrast, the
FuzzyScore1 or 2 modules positioned species (rows) with their scores of
variance equal to 1. A trait modality was positioned at the average of
species (rows) having the modality.
This operation is yet available in the ADE-4 version. Generate files
".flc1" and "xx.fll1" from the DDUtil : Add normed scores option after
having analyzed your table (MCA: Fuzzy correspondance analysis). Use the
Scatters module to superimpose files "xx.flli" and "xx.flc1" (equivalent of
FuzzyRowScore) or files "xx.flco" and "xx.fll1"(equivalent of FuzzyScore2).
Do not forget to set the scales on the largest values (l1 and c1
The last problem concerns the lines that linked modalities and rows. The
ScatterDistri module may help to view the above link. Use the ScatterDistri
: Frequencies option with the file "xx.fll1" and the xxF file that contains
the raw data table (after FuzzyVar : Read Fuzzy File). This procedure
results in as many graphs as the number of modalities. In a given graph of
the jth modality one finds the factorial map of rows. On the same graphs it
is possible to paste similar graphs resulting from the ScatterDistri :
Stars option processed on the same files as the ScatterDistri : Frequencies
option. This is a way to link centroids (modalities) with corresponding
rows. The parametrization of the Row & Col selection box enables the
selection of the modalities of only one trait.
To illustrate the reverse averaging principle just use the ScatterDistri :
Frequencies option with the file "xx.flc1" and xxFTR that contains the
transposed raw data table (after FuzzyVar : Read Fuzzy File and Files :
transpose). This procedure results in as many graphs as the number of rows.
In a given graph of the ith row one finds the factorial map of modalities
(weighted by their use at the ith row). On the same graphs it is possible
to paste similar graphs resulting from the ScatterDistri : Stars option
processed on the same files as the ScatterDistri : Frequencies option to
get lines that link modalities to a centroid for a given row.
According to Daniel Chessel (26/11/97), the fuzzyscore options were not
programmed again for the following reasons (which could be revisited
1) This kind of graphics are manageable only on Macintosh computer since
Windows and PC demonstrate poor graphical quality. Compatibility between PC
and Mac versions does not allow us to complicate to much the graphical
presentations of ADE outputs (especially the instability of the WMF format
is in contradiction with the very large application of the PICT format).
2) This kind of presentation was improper since a contant line thickness
represents only roughly a weighted link. Such graphs aimed at presenting
the underlying averaging but was weakly descriptive. The use of various
thickness can be just impossible for large data sets, therefore this
possibility was not imported in the ADE-4 version.
3) We gave priority to other methods that compete with FCA since this
latter technique does not allow a complete investigation of the so-called
fuzzy tables. Though it was a good introduction to the trait managing topic
as available in Freshwater Biology 31, it is far from solving all the
problems. For example, a species has a certain distribution for each trait.
For a given typology of modalities, a species has an average position for
each trait, and a global average position of the average positions for each
trait which cannot be seen by FCA. Moreover, the question of trait
independance or correlation is not directly addressed by FCA. This is why
in topic documentation 81 the so-called fuzzy table are treated using
If we add developments (in preparation) about trait biodiversity, you will
see that the question of analyzing trait tables is entirely open.
All the very best,
Université Claude Bernard - Lyon 1
43 Bd du 11 novembre
Bat 401C - 2ème étage
F-69622 Villeurbanne CEDEX
Tel : +33 4 72 43 13 63
Fax : +33 4 72 43 11 41
ADE-4 package is available on the Internet
This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:36:03 MET