...from the Microcomputer Power website.CANOCO 3.15 | <Previous | Next>
CANOCO 3.15 Read Me
CANOCO version 3.15 (September 1997)
*** Please read the CANOCO licence conditions first (file LICENCE) ****
*Changes of version 3.15 with version 3.14
*Changes of version 3.14 with version 3.12
*Demo runs in short
*Demonstration run of CANOCO
*Data formats for input files for CANOCO
*How to get the output displayed in the CANOCO manual (chapter 4)
*CANOCO comes in two sizes
*The new features in CANOCO 3.1 compared to CANOCO 2.x
*Message to users of the test versions labelled CANOCO 3.0:
*Hint in case you dislike to add C:\CANOCO to your PATH in the AUTOEXEC.BAT
*Installation on floppy disk
*Message to users of MS-DOS 3.2:
*DOS Error messages:
*Order information for more licences/site licenses
At present (November 1994), CANOCO version 3.14 for MS-DOS consists of the
- three 5.25 inch 360 Kb diskettes or two 3.5 inch 720 Kb diskettes or
a single high density diskette
- the manual of CANOCO 2.1
- update notes CANOCO 3.10
- the booklet with reprints "Unimodal models to relate species to environment"
- the manual of CANODRAW 3.0 (if ordered additionally)
- the manual of CEDIT (if ordered additionally)
Changes of version 3.15 with version 3.14
Version 3.15 of CANOCO uses the Strict convergence criteria of
Oksanen & Minchin (1997) Journal of Vegetation Science, 8, 447-454.
The bug in the subroutine reported by Oksanen & Minchin (1997) has been
repaired also. In fact, both tests on AZ3 in the subroutine
now use 'less or equal to 0.0'. This makes the routine order invariant.
A change in the subroutine SEGMNT ensures that samples with one species only
are skipped in calculation of the mean-squared deviation.
I thank Mark O. Hill for pointing out the improvement in the logic of the
non-linear rescaling in subroutine SEGMNT.
The changes in SMOOTH:
both changed to
The change in SEGMNT:
C SUGGESTED BY MARK HILL Decorana manual 1979 p 28 which says:
C no information about the average mean-square deviation
C is deemed to accrue
C THEREFORE SAMPLES WITH ESSENTIALLY ONE SPECIES ARE NOT COUNTED
C IF(SQCORR.GT.0.9999) SQCORR=0.9999
IF(SQCORR.GT.0.9999) GOTO 50
*Changes of version 3.14 with version 3.12
Version 3.12 is the version that has the widest distribution.
Version 3.13 nver existed!
Version 3.14 was an update repairing specific problems that were reported
to occur in version 3.12
A problem with the permutation test in very small samples has been repaired
A problem with the option 'more ordination axes' in canonical methods
*Demo runs in short
: Type, preferably in an empty directory, :
: CANOCO Press many times <RETURN> :
: Gives ordination of the species data (file DUNE.SPE) :
: Solution in file CANOCO.SOL :
: Yours answers are recorded in CANOCO.CON :
: CDLITE :
: Displays the ordination on the screen :
: Select CANOCO.CON as File with CANOCO answers :
: Select Biplot :
: CEDIT :
: Edits data files for CANOCO :
: Select DUNE.ENV as file to read :
: Press many times <RETURN> :
: For Help in the editor, type H after Edit? :
: Use PrintScrn to print the Help screens :
: To leave the editor type Q after Edit? :
With CANOCO you do ordination. To run CANOCO simply type CANOCO.
If you receive an error message at this point, see below under Error
messages or try to carry out a demonstration run.
With CANODRAW LITE (CDLITE) you can display ordination diagrams on the screen
and various other output devices. To run CANODRAW, type CDLITE.
See also file README.CDR. CANODRAW comes in addition to the program
CANOPLOT, which produces lineprinter diagrams only. To run CANOPLOT,
With CEDIT you prepare data files for input to CANOCO and
CANODRAW. To run CEDIT, type CEDIT. CEDIT is also very handy for
handling the input data for other programs that use or can use input
in Cornell Condensed Format, e.g. TWINSPAN. With CEDIT, you can edit,
transform, condens, decondens huge sparse data matrices. CEDIT as
supplied with CANOCO is a restricted version of the full program,
available from the author Onno F. R. van Tongeren (see file
*Demonstration run of CANOCO
For a demonstration run of CANOCO, simply type CANOCO <return>
and answer the questions by pressing <return> at each question. The
analysis then carried out is a canonical correspondence analysis of
the "Dune meadow data" (see the Manual, pp 10-12 or "Unimodal models
to relate species to environment": page 78). You may obtain a
printplot by typing CANOPLOT and pressing <return> at each question.
Alternatively, type CANODRAW to invoke the fully fledged graphics
This demonstration is made possible by virtue of the file CANOCO.INI.
This is a normal text-only file which specifies the default file
names and the default analysis to be carried out. To get the full
advantage of the file CANOCO.INI for your own analyses, copy
CANOCO.INI to the directory where your data are and edit CANOCO.INI
with an editor or a word processor; in particular, change dune.spe
and dune.env to the names of your data input files for CANOCO. (If
you used a word processor, make sure you saved the file as a
text-only or DOS-text file). Later on, if you know more about the
details of CANOCO and of ordination, you may wish to change more
defaults. But, before you can run your own analyses, you need to
learn which types of files CANOCO accepts as data input. This is
described below and in more detail in the manual. The file CANOCO.INI
also contains information to initialize CANODRAW and CEDIT.
CANOCO overwrites the output file or the solution file of a previous
run if the same name is specified.
*Data formats for input files for CANOCO.
CANOCO accepts three different formats of data files (example files in
(1) Cornell condensed format and CANOCO condensed format, e.g. file DUNE.SPE
(2) CANOCO full format, e.g. DUNE.ENV, BACI.SPE
(3) Free format, described below, see example file FREE.DAT.
Data on species, covariables and environmental variables may be in either of
the three formats. Note, however, that CANODRAW expects that the
species data are in CANOCO Condensed format and the environmental
data are in full format. If your files have a different format than required
for CANODRAW, type CANODRAW -Q -S, or transform the data files to the
appropriate format with the program CEDIT.
In addition to the CONDENSED format and FULL format described in the manual
of version 2.1, CANOCO version 3.1 allows free full format.
FREE FULL FORMAT:
Line 1: title
Line 2: keyword 'free' or 'FREE' starting in position 1 to 7.
Line 3: number of species (variables), number of samples (units)
Line 4: start of data.
The data of a sample may occupy more than one line, but each new sample must
start with a new line. Species and sample names are optional. Their place
starts at the first line after the data values (i.e. no blank line or "ending
Example of free full format:
example line 1 of free full format
2 1 0 9
0 8 2 4
3 1 8 0
0 0 0 1
4 0 1 8
At the end of the data, names of species and samples may be given as described
for the other formats. If the names are missing, CANOCO gives them default
SPEC 1, SPEC 2 etc for variables in the species file
V 1 , V 2 etc ,, covariables and environmental variables
SAMP 1, SAMP 2 etc
In Condensed and FULL format files, names are allowed to be missing as well
in version 3.1.
When an error is detected when reading data, CANOCO gives the sample number
at which the error is detected. In CANOCO 2.x, empty samples (zeroes only in
full format files) were discarded. In CANOCO 3.1, such samples are retained
in PCA/RDA analyses.
*How to get the output displayed in the CANOCO manual (chapter 4):
The Dune Meadow data are also used in the manual to explain the output of
CANOCO. However, the output that you obtain(ed) in the demonstration analysis,
differs from the output displayed in chapter 4 of the manual. The change is
due to another default choice of scaling of the ordination as explained in the
Upgrade Notes. If you wish, you can easily generate output that is more
similar to that shown in chapter 4 by using the "answer file" TABLE33.CON (the
version 3.1 equivalent of Table 3.3 in the manual). In the CANOCO directory,
CANOCO (the first CANOCO screen appears)
1 (yes, you wish input from file)
TABLE33.CON (No, not from ANSWER.CON, but from TABLE33.CON)
CANOCO will read the answers from the file TABLE33.CON. What is being read as
an answer to which question, is displayed, but if you have a fast PC, you will
not be able to read anything until CANOCO stops to display the correlation
matrix of the environmental variables. Compare with Table 4.1 in the manual.
The results in Table 4.2 are displayed more efficiently in the "Summary of
the ordination", after three times pressing return. Subsequent screens will
show the Tables 4.3, 4.5-11. The output is also listed in the output file
DUNE.OUT and solution file DUNE.SOL.
(If you scrutinize the output carefully, you will spot a difference with Table
4.3 (the species made passive in Table 3.3 are not present in the file
DUNE.SPE) and also with Table 4.6 and table 4.7. The latter does not need to
bother you at this moment; the explanation requires more knowledge of
regression than is needed for a default analysis of CANOCO. But, you may
notice that in these tables the variables 4 and 7 (HAYFIELD and SF) are
missing, whereas in DUNE.SOL the variables 5 and 10 have zero scores; see Q20
on page 25 of the manual and pages 58-59 of Jongman et al, 1987).
EX61a.CON and EX61b.CON are CANOCO answer files that can similarly be used to
reproduce the analyses on which the example in section 6.1 of the manual is
based. (Both analyses take some time to run because of the Monte Carlo
permutation test). By answering EX61a.CON instead of TABLE33.CON above, you
obtain the analysis of Fig. 6.1. EX61b.CON is a partial canonical
correspondence analysis. It should be noted that the significance values (P-
values) obtained may differ slightly from those reported in the manual
because (1) less permutations are used, (2) other seeds are used (see Q39 in
The extension .CON in the CANOCO answer file is not obligatory, but useful
when using CANODRAW for making ordination diagram. The historic origin of the
extension .CON stems from the word CONSOLE which is an old fashioned word for
the terminal screen. The answer file records what is typed at the console.
If you do not wish to see the screen output, when using CANOCO with .CON
files, use CANO.BAT, e.g. type at the DOS prompt
The output is in the files EX61a.OUT and FIG61.SOL, as specified in EX61a.CON.
*CANOCO comes in two sizes:
The 'small' CANOCO.EXE, which is unpacked in the default directory
(C:\CANOCO), requires at least 420 Kb free memory. A mathematical
coprocessor is optionally. This program is dimensioned for 500
samples, 500 species, 100 covariables and 58 environmental variables.
The total data space is 45000 reals.
The 'big' CANOCO.EXE, which may be unpacked in the subdirectory ..\1000
(default C:\CANOCO\1000), requires at least 580 Kb free memory. Use the
DOS-command CHKDSK or the utility MAPMEM to check how much free
memory your PC has (DOS 4 user can use MEM/PROGRAM). If the free
memory is less, remove resident programs (which are probably loaded
in the AUTOEXEC.BAT file). A coprocessor 80x87 is required. This
program is dimensioned for a maximum of 1000 samples, 700 species,
100 covariables and 75 environmental variables. The total data space
is almost 80000 reals (to be precise: 78800).
*The new features in CANOCO 3.1 compared to CANOCO 2.x:
- forward selection of environmental variables (with testing if desired; see
FORWARD.CON for automated testing)
- regression/ordination diagnostics for samples and species
- redundancy analysis with a statistical weighting of species
- percentages variance accounted for in CA (and CCA)
- Collinear environmental variables (e.g. the K-th dummy variable of a nominal
variable with K classes) are kept in the analysis. The default analysis thus
yields CENTROIDS for all classes instead of K-1 classes. In CANOCO 2.x this
required an extra analysis with passive environmental variables; compare Table
3.3 of the manual with TABLE33.CON.
- data input in free format (see below). The error messages when reading all
types of input are improved.
- principal component scores of species adjusted or unadjusted for species
variance. The new default is 'adjusted', which means that species scores are
exactly the correlations of species with axes if the scaling of a correlation
biplot is chosen (the former covariance biplot).
- Hill's scaling is no longer the default in CA/CCA. Instead, the default is
the biplot scaling, which is easier to interpret (scaling to 1 instead of to
1/(1-lambda) and to lambda instead of to lambda/(1-lambda) ).
- screenwise output of most output. CANOCO can present screenwise output if
an ANSI screen code driver is installed on the computer (e.g. DEVICE=ANSI.SYS
in the CONFIG.SYS of a MS-DOS computer). After the message at the bottom of
'Press RETURN for more, S to Skip, Q to Quit page-mode'
the user can press <RETURN> to continue, S <RETURN> for skipping page-mode
for the current output item, or Q <RETURN> to quit the page-mode for the
entire CANOCO session. Subsequent output starts from the top of the screen.
- choice between short and long dialogue.
In the short dialogue, one can delete samples, covariables,
environmental variables and choose a transformation for the species data. The
analysis can be continued in the usual way with passive analyses, Monte Carlo
statistical tests, etc.
- you may specify your own defaults in a file CANOCO.INI. This is especially
convenient when using the short dialogue. CANOCO.INI must be in the current
directory or in C:\CANOCO. A CANOCO.INI file in the current directory
overrules the file in C:\CANOCO. The format of the example file must not be
changed. All elements between *CANOCO and *ENDCANOCO are required in that
order. The defaults that give results similar to version 2.x are given in the
file VERSION2.INI. To enable these defaults, use the copy command: COPY
c:\canoco\version2.ini c:\canoco\canoco.ini For ease of comparison, the
defaults for version 3.1 are in VERSION3.INI. The file CANOTAB.INI differs
from VERSION3.INI in specifying a TAB to separate values in the solution file.
-more than 4 ordination axes may be extracted by asking for More analyses with
current data (instead of Stop).
-environmental variables can be turned into covariables after the question
which asks for more analyses with current data.
-The methods used for the Monte Carlo test have been changed. See the Update
Notes. The old and new method yield equivalent results in simple situations
as overall tests without covariables and (with permutation under the null
model) tests with a conditioning on all covariables (randomized block
- in Monte Carlo testing, special permutation schemes have been added: for
time series, for samples on a line transect, on a grid and for repeated
- the former machine readable copy is now termed 'solution file for CANOPLOT
or other program'. The numbers in that file can be in two formats (see
CANOCO.INI). A decimal format which reports the ordination scores as they are,
and a non-decimal (integer) format which reports the scores in whole numbers
with the original scores multiplied by the MULTIPLIER (1, 10, 100 or 1000).
In version 3.1, the decimal format is the default (see VERSION3.INI), whereas
in version 2 the integer format was the default. With the decimal format, the
user can specify, in the CANOCO.INI file, symbols to separate numbers, to
enclose names and to signal the end of a set of scores. CANOPLOT does accept
both types of format of the solution file. CANODRAW accepts the decimal format
- in the solution file for CANOPLOT, a column headed N2 has been added to the
species and sample scores. In unimodal methods this contains, for samples,
Hill's (1973, Ecology 54: 427-432) diversity measure N2. N2 is the effective
number of species present in a sample. For species, N2 is defined analogously
and means the effective number of occurrences of a species. N2 is the inverse
of the Simpson index. In linear methods, N2 is arbitrarily assigned the value
- The MS-DOS redirection symbols > and < work now properly with the answer
file CANOCO.con. Before using such file, rename it, for example by RENAME
Some example CON files are on the distribution disks.
You can run CANOCO from the DOS prompt by
1. CANOCO < RDA.CON > SCREENDUMP (do not read from CANOCO.CON as CANOCO will
make a new CANOCO.CON also)
or, to get the full screenoutput, type
2. CANOCO > screenfile
or, for no screenfile,
3. CANOCO < rda.con > NUL
The advantage of the fourth way of invoking CANOCO is that, if the answer file
is exhausted, CANOCO returns to the screen for further input. Further answers
given interactively are written to CANOCO.CON. Therefore, always rename
CANOCO.CON before using it as input for CANOCO.
In a CON file, files names can be specified as FROM INI. The file name is then
taken from the CANOCO.INI file. If the file name equals S, the default file
name is Skipped, and the analysis proceeds without the file of inquiry.
The third way is the quickest. To make this way work with programs like
PCSHELL from PCTOOLS or DESQVIEW or WINDOWS, use CANO.BAT, which has a single
line (CANOCO < %1 > nul). By specifying CANO.BAT as program information and
*.CON files as specific for running CANOCO, you may integrate CANOCO into the
environment you are accustomed to.
- the output option for T_VALUES in the ordination results, now yields
coordinates for species and environmental variables that form a biplot which
approximates t-values of the regression of species on to the environmental
variables (ter Braak, 1990). This option yields output only for RDA and for
CCA with Hill's scaling disabled.
- In PCA/RDA and decimal format for the file for CANOPLOT, the projection of
the true origin (ECOLOGY, 64: 454-462 eq (3)) is given as last sample in the
sample scores on the file for CANOPLOT under the name of ORIGIN.
- In the file CANOCO.PUN (the default name), CANOCO writes an
environment-by-species table. This table contains:
in CA/CCA/DCA: weighted averages for standardized environmental variables
in PCA/RDA: correlations between species and each environmental variable,
when species are centred and standardized, and similar
(values proportional to) covariances, when species are just
- more flexible memory management, whence more data can be analyzed.
*t-value biplot (T-biplot in CANODRAW)
The relation between multivariate multiple regression and redundancy
analysis has been described in the CANOCO manual (section 7.3),
Ter Braak (1990) and Ter Braak & Looman (1991). The species scores
and Canonical coefficients for environmental variables give a biplot
of the matrix of regression coefficients of the multivariate regression.
The approximate t-ratio's of these regression coefficients are displayed
in a biplot of
StBi: Species coordinates for t-value biplot
EtBi: Environmental coordinates for t-value biplot
in the solution file produced by CANOCO.
This biplot has been described in the Update Notes in section 10 and
in Ter Braak (1990).
One interpretation aid should be added. It is possible to
indicate in which region the species lie that react significant and
positive to a particular environmental variable. By elementary
geometry, this region is a circle with as middle line the line
segment joining the EtBi-coordinate with the origin. The species with
StBi-coordinates that lie inside the circle have, in the
approximation, a significant, positive regression coefficient on that
particular environmental variable. By mirroring the circle in the
tangent at the origin, we find the region of species with
significant, negative regression coefficients. Ter Braak & Looman
(1991) present an example. The possibility of the above interpretation
aid was first noted by H. van Dobben (IBN-DLO, Wageningen). The circles
are therefor called van Dobben circles in Ter Braak & Looman (1991)
Ter Braak, C. J .F. (1990). Interpreting canonical correlation analysis
through biplots of structural correlations and weights.
Psychometrika, 55: 519-531. (also treats [partial] redundancy analysis)
Ter Braak, C. J. F. & Looman, C. W. N. (1991). Biplots in
reduced-rank regression: a public health example. Biometrical Journal,
For a minimal installation, e.g. on floppy disk, see at the end of this file.
On a hard disk system, make a directory, preferably, C:\CANOCO and
make this directory the current one, for example by typing
and invoke the files INSTCAN, INSTDRAW and INSTCED for expanding the
files of CANOCO, CANODRAW and CEDIT, respectively.
Before invoking CANOCO, add the directory C:\CANOCO to the PATH in your
AUTOEXEC.BAT file. CANOCO and CEDIT use ANSI codes for output to the
screen. So check your CONFIG.SYS file to see whether there is a line
or your AUTOEXEC.BAT for a line like
which invokes an ANSI.COM file. Go to the CANOCO directory and type
CHECKPAT or ask help from a friend if you are not familiar
with these DOS features.
If you choose to install CANOCO, CEDIT and CANODRAW on another
directory, check the file CANOCO.INI and change the references to
If your licence allows installation on a network, the users
should be recommended to copy the CANOCO.INI file from the network to their
current [WRIE-enabled] directory or their C:\CANOCO.
*Hint in case you dislike to add C:\CANOCO to your PATH in the AUTOEXEC.BAT:
If you use CANOCO and CEDIT only now and then, you may dislike to
extend your path because is long already. This is a trick to create a
PATH with 'a directory-on-the-fly'. It works from DOS 3.x onwards.
Add to CONFIG.SYS a line like
Add in AUTOEXEC.BAT
SUBST F: C:\
When you now wish to use CANOCO or CEDIT start with typing
Check the result with DIR F:
Of course, this trick may be useful for other programs you use irregularly.
*Installation on floppy disk
This presents no problem on 3.5 inch disks or on high density 5.25
inch disks. For 5.25 inch 360 Kb drives, you need the help of a
friend with a PC with hard disk to expand all files. For running
CANOCO, you need only CANOCO.EXE and CANOCO.INI. For running CEDIT,
you need CEDIT.EXE, CANOCO.INI, CEDIT.MSG, CEDIT.MAC. For running,
CANODRAW, you need CANODRAW.EXE, CANOCO.INI and the appropriate *.BGI
*Message to users of MS-DOS 3.2:
The Microsoft Fortran 5.0 manual (Advanced Topics, p. 15) reports that there
is a bug in this MS-DOS version. You may have problems using the math
coprocessor. There exists a Patch (path87.exe) for this problem to be asked
for from Microsoft. An intermediate solution might be to disable the math
coprocessor by setting a DOS environment variable:
SET NO87=Use no math coprocessor
The problem is solved from MS-DOS version 3.21 onwards.
*DOS Error messages:
Bad command or file name
Cannot find CANOCO.EXE
Please enter new program specification:
Both errors indicate that CANOCO.EXE is not in the current path. Either change
directory (CD) to the directory where CANOCO.EXE is or add this directory to
the path in your AUTOEXEC.BAT file. The program CHECKPAT can add the CANOCO-
directory to your path. First, make the directory with CANOCO.EXE the current
one (using CD) and type CHECKPAT.
Cannot execute CANOCO.EXE
Program too big to fit in memory:
You probably answered Y to the question CANOCO.EXE OVERWRITE [Y/N].
This gives a version of CANOCO which requires 580 Kb free memory. You
probably loaded a number of memory resident programs in your
AUTOEXEC.BAT file, so that you have too few free memory. With the
DOS-command CHKDSK to can find out how much free memory your system
has. Delete CANOCO.EXE (Essential!!!) and install CANOCO again,
answering N to all OVERWRITE questions. This gives a CANOCO.EXE,
which requires 405 Kb. If you wish to run the version of CANOCO with
larger dimension, please change your AUTOEXEC.BAT file. You may also
add to your CONFIG.SYS file the line STACKS=0,0 or lower the number
of buffers and files, e.g. to 10 and 20, respectively. On
386-machines, you can buy programs to increase the amount of memory
available for DOS, e.g. QEMM-386 or DESQVIEW-386 from Quarterdeck. If
you are using the command line editor CED, invoke it in a minimal
configuration, e.g. CED -b128,128,128,128,128,128 or buy Norton's
NDE, which requires on 3 Kb.
Floating point not loaded:
You answered Y to the OVERWRITE question during installation.
The version of CANOCO.EXE then obtained, requires a mathematical 80x87
coprocessor. Delete CANOCO.EXE (Essential!!!) and install CANOCO again,
answering N to all OVERWRITE questions.
If you see such strange strings appear on your screen, your computer
does not accept the ANSI codes for screen output. See notes on
manual installation about ANSI.SYS and ANSI on. You can also try to
run the program CHECKPAT in the CANOCO-directory, or do the whole
installation of CANOCO again.
*Order information for more licences/site licenses
CANOCO program can be obtained from:
In Europe: Campus Software, Vadaring 29, 6702 EA Wageningen, the Netherlands
Outside Europe: Microcomputer Power, 111 Clover Lane, Ithaca, NY 14850, USA.
phone: (607) 272-2188
From them, you can obtain a free version of CANOCO for demonstration for your
friends. CANOCO is portable to many other computer makes, e.g. it runs
under VAX/VMS and IBM mainframe computers. There is also a version available
for the Apple Macintosh. Please, inquire at Microcomputer Power. An OS/2
version which can analyze up to 2000 samples, is also available.
The DLO-Institute for Forestry and Nature Research (IBN-DLO) and its
researchers are gratefully acknowledged for their support of the
CANOCO project. I would like to thank Onno F.R. van Tongeren for
writing CANOPLOT and CEDIT and for his permission to supply CANOPLOT
free of charge with CANOCO, and Petr Smilauer for writing CANODRAW.
These programs add much to the ease-of-use, power and flexibility of
If the text just ran over your screen, type now:
type readme | more
This page was last rendered at Ithaca, NY Monday, August 24,
1998 2:26 AM