ANALYSING STABILITY OF CANOCO ORDINATIONS

STABLTY.EXE is a program for automatic analysis of stability of ordination solutions with CANOCO 3. It is a utility program that

All these operations can be performed with a single command STABLTY in a MS-DOS window. Stablty is compiled with djgpp (MS-DOS port of GNU C or gcc), and it needs a memory manager to run. The most common nowadays is Windows dpmi, which you can use by running Stablty in the MS-DOS prompt of MS-Windows.

CONTENTS OF SHUFFLE.ZIP

This pk-zipped package contains the following programs:

Other files in this package include:

WARRANTY: ABSOLUTELY NO WARRANTY. I DON'T GUARANTEE THAT THESE PROGRAMS WORK OR THAT THEY CAN BE USED FOR ANY PURPOSE.

USING STABLTY.EXE TO ANALYSE CANOCO STABILITY

Some simple analyses of CANOCO instability can be performed automatically with the help of Stablty.exe. For this purpose you need the programs CANOCO v3, Cepshufl v1.4 and Solcomp v1.4. All these programs must be on the path or in the same directory were you run stablty.

WARNING: It is best to make safety copies of the programs in a different directory before running Stablty.

Program Stablty asks for some details of CANOCO analysis. The program writes a corresponding CANOCO instruction file (canoXXXX.con), a batch file to run shuffling and comparison (stblXXXX.bat), complete monitoring file from Solcomp (moniXXXX.txt), and outputs summary of the results on the screen. In addition, the program produces temporary files called shufXXXX.* and targXXXX.* (several types). In all these files XXXX is numeric identifier, so that there won't be conflicts with existing stable file names. However, the program does not check for the existence of temporary files (shufXXXX, targXXXX), and these files will be overwritten. Therefore it is best to inspect that there are no necessary files with those names. If you do, either rename your files or move them to a different directory.

Program Stablty runs one CANOCO with the original data input order. After that it runs a number (default 16) of Cepshufl-> CANOCO -> Solcomp cycles, and saves all comparison results in moniXXXX.txt. Then it scans the moniXXXX.txt file and reports some summary statistics.

Interpretation of Stablty output

Stablty reads the monitor output file moniXXXX.txt and prints the following summary results on the screen:

The following statistics are given separately for species and sites.

Command line parameters

The user interface is very simple. The program Stablty asks only for a some few parameters of CANOCO that I think are interesting, and then writes a CANOCO instruction file (canoXXXX.con) from which CANOCO reads its parameters. In addition to the CANOCO parameters, you can change the number of shuffling cycles, by calling Stablty with a numeric value on the commandline. If you want a comparison based on 10 shuffling cycles, invoke Stablty with the command:

STABLTY 10

If you want to analyse and summarize an existing moniXXXX.txt file, invoke Stablty with parameter 0:

STABLTY 0

If you want to write the batch file (stblXXXX.bat) and CANOCO instruction file (canoXXXX.con), but do not want to execute them immediately, invoke Stablty with a corresponding negative parameter. To produce a default number of shuffling cycles without executing the file, use

STABLTY -16

This property is useful if you are not satisfied with the Stablty user interface, and want to edit the CANOCO instruction file before running comparisons.

If you want to repeat the same shuffling for the same data set, you may give the random number seed as a second command line parameter after giving the number of cycles. The seed must be a positive integer below ca. 1000000, e.g.,

STABLTY 16 314159

Modifying CANOCO parameters

If you want some parametrization of CANOCO that is not supported by the Stablty user interface, it is best to modify the CANOCO instruction file canoXXXX.con and after that run the batch file and comparison. CANOCO normally ouputs the file with the answers to the questions (usually called CANOCO.CON), where you can more easily see the options used and the modifications needed. You can take the following steps:

  1. Modify the current instruction file canoXXXX.con.
  2. Delete the corresponding old Solcomp monitor file moniXXXX.txt, so that the new output will not be appended to the old file. If you have invoked Stablty with negative parameter, you can omit this step, since moniXXXX.txt does not exist.
  3. Run the batch file stblXXXX.bat with writing in the MS-DOS prompt: STBLXXXX
  4. Analyse the new monitor file moniXXXX.txt using the command STABLTY 0

and give the monitor file number XXXX when the program asks for the monitor file identifier.

Options which concern only some specific species or sites cannot be modified by this method, since the ordering of species and sites is unknown after running Cepshufl. The only way of handling these cases is to edit data files so that there is no need to omit or passivize specific sites or species. However, it is possible to invoke Cepshufl with switch -e which requests sorting for sites only. Stablty invokes shuffling of environmental data sets with this switch, and so you can edit individual environmental variables with CANOCO instruction file.

It is possible to analyse the stability of constrained ("canonical") ordination as well. However, there is need to be especially careful, since errors are easily made when handling two data sets together. If the results from a constrained ordination are much worse than those from the corresponding unconstrained ordination, it may be best to seek first errors in own definitions before dumping the constrained analysis. Both the community data file and the environmental variable file must contain the same sites in the same order. If there are sites without environmental data, they must be removed from both files explicitly before running the analysis.

Good luck!

Jari Oksanen 16/9/97

AN EXAMPLE ON THE STABLTY SCREEN OUTPUT:

==================================================================
SOLCOMP 1.4 (Mon Sep 15 14:22:12 1997) - J.Oksanen & P.Minchin
DUNE MEADOW SPECIES DATA (M. BATTERINK AND G. WIJFFELS, 1983)
Method DCCA, Scaling -1 --- 30 species, 20 sites
===Eigenvalues by axis:
Axis 1: 0.3391
Axis 2: 0.1148 .. 0.1149 *** UNSTABLE EIGENVALUE
Axis 3: 0.0410
Axis 4: 0.0072 .. 0.0097 *** UNSTABLE EIGENVALUE
===Species ordination:
Maximum non-rotated rmse (sd): 0.8823
Maximum rmse (sd) by non-rotated axes:
0.0001 0.0335 0.0227 0.8817
Worst diagonal elements of the Procrustes rotation matrix:
0.9980 0.9969 0.9799 0.9810
===Site ordination:
Maximum non-rotated rmse (sd): 0.9853
Maximum rmse (sd) by non-rotated axes:
0.0001 0.0074 0.0058 0.9853
Worst diagonal elements of the Procrustes rotation matrix:
0.9860 0.7072 0.4671 0.1639
Program wrote the following files:
moni0014.txt - Complete comparison output
stbl0014.bat - Batch file to run the comparison
cano0014.con - Canoco instruction file