RLQ cooking

From: Sylvain Doledec (sylvain@biomserv.univ-lyon1.fr)
Date: Mon Aug 10 1998 - 14:35:59 MET DST

Dear John Stark

Thanks for being interest in our work. You point out some trouble you had
in processing RLQ analysis. Here are some advises.

>I'm trying to understand the RLQ analysis by running it using the
>AVIURB data set as presented by Doledec et al. 1996 (Environmental
>and Ecological Statistics 3:143-166). I am using the WIndows 95

All figures and computations of the paper were made using a version of ADE
for PowerMac. We also used one quickbasic programs (Histo1Categ) that
helped to make fig8a and b of our paper. But as indicated in the module
documentation there are many ways of interpreting RLQ graphically. May I
suggest you to have a look to that documentation?
All figures were reorganised using a drawing software (ClarisDraw for Mac)
to make better printouts. So there is some further job after ADE produced
the graphs. As a result it may happen that the raw graphs you get with ADE
have not exactly the same shape as ours in the paper.

>With reference to the above paper, I have managed to figure out where
>all the numbers in Table 4(b) come from (although I'm not sure that
>I understand what they all mean).

All these numbers are scattered in various listing resulting from the
computations. They represent measurements of total variance, variance
associated to axes (eigenvalues), covariance and correlation. You can get
the values by making separate analyses of the 3 tables (R, L, and Q) and
then co-inertia analysis of pairs of tables (R-L and Q-L) and finally you
can get the numbers with RLQ analysis. So at least five analyses are
necessary to get the numbersŠ Sorry that they are not directly processed
from one module!

>I have also produced the
>equivalent of Fig. 3 except that I have been unable to determine how
>to get the lines joining the sites (black squares) to the appropriate
>environmental variable code ( a or p for 10 variables, and a - h for
>veg. cover). I assume it has something to do with the "Centers File"
>as requested by the ScatterClass.Stars module. I can draw lines from
>the origin to each site but not between the sites and the
>environmental variables. Is there a file produced in the analysis
>that tells me which sites connect with which environmental

To get the appropriate code you always can change the content (but not the
number of line) of xx.123 files that automatically include labels such as
(1a, 1b, 2a, 2b,Š11h). You also can use your own label (text file). No
problem with ScatterClass.Stars module in the new version of ADE either for
PowerMac or PC. Do not forget to use the right weighting file for making
the figures since each site was weighted by its margin frequency in the
avifauna table (.fcpl) and each species in the trait table (fig.4) was
weighted by its overall frequency (.fcpc).

>I note in the ADE documentation the use of the ADE.old and Quickbasic
>modules - presumably only on the Apple Mac. Is the Windows 95
>version limited by comparison with the Mac version?

The ADE.old modules do not run anymore on MAC since the system version
changed. A lot of Quickbasic modules were translated into C but not all.
Histo1Categ (for doing Fig.8) is not available but RLQ results in code for
the average position of modalities of traits. So use these codes to have
the results. Tab-MeanVar (Fig. 7) was translated. So you can interpret RLQ
even if not all graphs are available.

>How does one produce figures like Fig 6 - 9 using ADE? If the
>graphics require external programmes, that's fine, but what files
>contain the data that are plotted, and how are these created by
>the analysis?

This is all explained in the topic documentation (Appendix), which I
apologize is only available in French. However in combination with the
module documentation (see for example the graphics at the end of
RLQ:co-inertia analysis option p.7) of RLQ you may succeed.

All the very best for ADE-lister at work today.

This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:22:00 MET