Detrended Correspondence Analysis

Robert Knox knox at spruce.gsfc.nasa.gov
Tue Jan 18 13:04:04 EST 1994


In article 4947 at dflora.cuug.ab.ca, peter at dflora.com (Peter D. Wilson) 
writes:
>Posting for a friend:
>
>I have been asked to trace some software which performs 
>"Trended Correspondence Analysis". It is to be used in botany and is  
>believed to have been created by a botanist. Any help would be  
>appreciated.
>
>- peter
>-- 
>peter d. wilson                                                         
>email: peter at dflora.com  (NeXTMail OK)            
>Voice: (403) 283-9743  Fax: (403) 283-0036                            
>     Independent NextStep Developer                                           

The technique is "Detrended Correspondence Analysis," where detrending 
refers to subtracting a smooth non-linear dependence among the axes, 
characteristic of correspondence analysis solutions for data where 
there is lots of species turnover amoung the sampling locations.  The 
most popular computer code for calculating the solution is a Fortran 
program called DECORANA, by Mark Hill.  Along with various public and 
'at cost' implementations, there are supported commercial versions.  
Most owe something to Hill's algorithm.  I append a list of old 
sources.  I don't know which of them are still valid contact points.

If the data lend themselves to simple a priori hypotheses about what 
factors control the patterns of variation, I'd recommend*:

  CANOCO (Vegetatio 75:159-160.):
     Cajo J.F. Ter Braak
     Agricultural Mathematics Group
     Box 100, 6700 AC Wageningen
     The Netherlands

The North American distributer:

     Microcomputer Power
     111 Clover Lane Dept. 16
     Ithaca, NY 14850 USA 
     (607) 272-2188

Your friend might want to look at an article by Mike Palmer in the December 
issue of _Ecology_:

  Palmer, MW. 1993. Putting things in even better order: the advantages 
     of canonical correspondence analysis.  Ecology 74(8): 221-2230.


According to Lee Belbin, Dan Faith, and Peter Minchin, we should all be 
using modifications of global nonmetric multidimensional scaling (NMDS) for 
these problems.  The difficulty remains finding fast, robust, and cheap 
software for doing NMDS in ways appropriate for vegetation data.  Readily 
available NMDS code (e.g., ALSCAL) uses a least squares algorithm that 
negates most of its advantages in ecological situations where principal 
components analysis and correspondence analysis produce highly curved or 
involuted solutions.

Any bionet.plants readers have suggested sources for KYST-style NMDS code?

- Bob Knox
- Biospheric Sciences Branch, Code 923
- NASA/Goddard Space Flight Center, Greenbelt, MD 20771 
- knox at spruce.gsfc.nasa.gov

------------------------------------------------------

  PC-ORD:
     Holcomb Research Institute
     Butler University
     4600 Sunset Ave.
     Indianapolis, Indiana 46208 USA

  ECOSURVEY:
     T.J. Carleton
     47 Waller Avenue
     Toronto, Ontario M5S 1B8
     Canada

  PATN (large system of programs for numerical taxonomy and
     classification):
     Lee Belbin
     CSIRO Division of Wildlife and Rangelands Research
     P.O. Box 84, Lyneham, A.C.T.
     2602, Australia

  DECORANA (public domain):
     Bradley Smith
     Forestry Sciences Laboratory
     Oregon State University


* These recommendations and suggestions do not constitute an endorsement 
of any product by the US Government or NASA or . . .



More information about the Plantbio mailing list