Thanks to everyone who weighed in on my question about community analysis 
software.  All the responses are pasted below.  There were lots of votes for R 
(with the Vegan package), and for PC-ORD.  It seems that for quick results and 
ease of use, PC-ORD is better, while R takes time to learn, but can be used for 
almost anything.

By the way, one person suggested Ginkgo (which I'd never heard of), which is 
part of a group of vegetation analysis programs called "VegAna".  It's 
available for free from a research group in Spain, and it runs in Java, so it 
works on both Mac and PC.  It is a quick download, and has a menu driven GUI 
and looks like it will produce results quickly.

Others suggested:  Primer, PAST, CAP, FATHOM (for MATLAB).

I think the best way to go is, as one person put it: "to compromise time and 
funds if you can: Purchase PC Ord, get your work done now (and support ecology 
professor, not computer programmer), and start learning R assuming to shift 
towards it as your statistical needs and familiarity grows."


Thanks again,
Joe

=====================================================================
I've used both PC-ORD and R for various community analyses.  I think Canoco is 
kind of obsolete.  I find PC-ORD very easy to use, and pretty versatile.  The 
options are pretty flexible.  Also, the manual is pretty good.  I also very 
highly recommend McCune and Grace's Analysis of Ecological Communities book.  
In R, of course, you can theoretically do anything you could possibly dream of, 
but it's a lot of work to learn, and the documentation is spotty (since it's 
basically written by various volunteers).  So it can be tough to figure out 
what the functions are actually doing, or how to get all the results out, etc.  
I'd say it's definitely worth learning, but it may take some doing.  If you 
just want to get most of your standard analyses done, I'd go with PC-ORD.  You 
can let me know if you have more specific questions.  Good luck!

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

I haven't used anything else, so this recommendation might not carry 
much weight, but I'm a big fan of PC-ORD. It's highlight is automating 
the nonparametric iterative analysis (Nonmetric Multidimensional 
Scaling Ordination, MRPP to compare between groups, cluster analysis, 
indicator species analysis - those are just the tools I've used), but 
it also includes the more traditional tools like Detrended 
Correspondence Analysis, TWINSPAN, etc. There is no programming 
required, and use is pretty intuitive with dropdown menus (and the 
help files are excellent). Graphics are publication-worthy. The McCune 
and Grace text (Analysis of Ecological Communities) goes into great 
detail on all the mathematical details behind each analysis technique, 
but fortunately one doesn't have to understand all of that because 
they sum up with recommendations appropriate to community analysis and 
give examples that spell out how to implement the technique in PC-ORD. 
I'd say the only tricky thing is setting up your dataset, but that's a 
piece of cake once you're familiar with the structure required.

Users of PC-ORD are also set up to go right into using the relatively 
new HyperNiche, also put out by MJM software, as the interface is very 
much the same. HyperNiche is good for regression, including all the 
traditional regressions as well as Nonparametric Multiplicative 
Regression, which basically allows one to define species niche-space 
in terms of environmental variables, to evaluate areas of optimums for 
your response variable given a whole suite of predictors. (I used it 
to estimate tree ring age counts from various predictors and was 
convinced that it was a much more appropriate method that MLR, even 
though my dataset did conform to the MLR assumptions.)

-----------------------------------------------------------------------
I use R and sepecifically the 'vegan' package (short for vegetation
analysis), it does most of what PC-ORD does. Also the authors of the vegan 
package have prepared a really nice user manual/tutorial.

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

I expect this is similar to many of the responses you're receiving but I
think it comes down to how much delay you can afford before producing your 
results and whether you'd rather invest time or money. 

R, which I'm currently teaching myself, does have a bit of a learning curve but 
has near-comprehensive capabilities and is growing rapidly into the statistics 
tool that can do anything.  It will expand your horizons
considerably and is pretty easy to pick up if you're already familiar with 
something like Matlab.

On the flip size, I've been using PC Ord for a few years and been very
pleased with its capabilities and ease of use.  One draw back is that I
don't believe it does Discriminant Analysis (I believe this is true for the 
newest version as well).  However, it gets top marks for tech support as I once 
reported a bug in the software, corresponded directly with the author received 
a patch within two days.

My recommendation would be to compromise time and funds if you can: Purchase PC 
Ord, get your work done now (and support ecology professor, not computer 
programmer), and start learning R assuming to shift towards it as your 
statistical needs and familiarity grows.

-----------------------------------------------------------------------
I have used PC-Ord, CANOCO, and R extensively, although R
for non-ordination analyses.  I would say without hesitation that PC-
ORD is the most user friendly and has the shallowest learning curve. 
It also has a very nice assortment of ordination alternatives and
would be the obvious choice for teaching.  On the downside, PC-ORD
requires a funky data setup that is a pain to continually monkey with
(unusual format in Excel and then use Save As to convert to a .WK1
file).  I'm not sure if this formatting issue has changed in newer
versions.

I've found CANOCO to have greater flexibility and options for
advanced analyses, e.g. principal response curves, redundancy
analysis, and many others.  CANOCO has traditionally been very
difficult for graphing the output (but PC-ORD is excellent at this).
While I haven't used R for ordination, this software requires a
fairly advanced knowledge of command line programming and even though
there is a graphical user interface, it wouldn't be easy for
students.  There is a lot of available support literature for R and
there are many, many contributed packages that make the command-line
stuff easier.  It's of course also free.

Ultimately, because I view CANOCO and PC-ORD as complementary
packages and am glad I don't own just one.  I'd probably recommend PC-
ORD for someone just starting out.  All of the multivariate analyses
I've published, however, have been done with CANOCO.   

-----------------------------------------------------------------------
The learning curve for R is pretty steep, but in the end it will give you a lot 
more flexibility (and it's free!). It could be weeks to months before you get 
your results though, if you take the time to learn it.

On the other hand, if you want to do your analyses right now, PC-ORD is very 
accessible and you'd be getting results pretty much right away. It has most 
basic community/ordination analyses built in with a menu format. I would look 
into what analyses you'd like to do and see if they're available in PC-ORD. If 
you decide to go that way, I also recommend the book Analysis of Ecological 
Communities by McCune and Grace (developers of PC-ORD).
(I did my dissertation analyses in PC-ORD and I liked it a lot but I did run 
into a few issues as far as flexibility goes. Now I'm trying to learn R and 
it's rough unless you're used to programming, but I can definitely see the 
benefits. Once you figure it out, you can ask pretty much any question of your 
data that you want.)

-----------------------------------------------------------------------
I have found CAP to be very easy to use (relatively of course)

And it is free!!!

Go to the web site of either the paper or the one in the user notes to get the 
program

-----------------------------------------------------------------------
I have experience using PC-Ord and R (but not Brodgar).  You do not have
to know R or anything like R (e.g. S Plus, SAS, etc) to use PC-Ord.  It
is a pretty good program with a nice graphical user interface, good help
files, and the learning curve is not steep at all.  I think that Brodgar
puts a graphical user interface on R (so that you don't have to do
anymore command line stuff), but I am not familiar with the types of
analyses that they offer.  PC-Ord has a lot of different analyses (and
you probably can't go wrong with it if you are interested in a package
with a nice GUI), but certainly not all that one might need as a
community ecologist, which is why I reluctantly made the switch to R a
couple of years ago.

-----------------------------------------------------------------------
You're hearing about some of the point and click programs, just thought I'd 
alert you to David Jones' FATHOM code for Matlab --

http://www.rsmas.miami.edu/personal/djones/matlab/matlab.html

Tons of great stuff there, but you need to know how to use Matlab minimally 
(copy in data, call the files, etc.).

-----------------------------------------------------------------------
I second the plug for R. It does take a while to get comfortable w/ it
but, once you do, sky's the limit. The vegan package is great and has
decent documentation (see attached). 

-----------------------------------------------------------------------
You might consider PAST.  It is free and Windows-based.  Here is a link to the 
download page: http://folk.uio.no/ohammer/past/

You can find documentation discussing PAST capabilities there.  It is easy to 
download and use.

-----------------------------------------------------------------------
R, on its own (well with a text editor to write the scripts; Emacs + ESS
is my personal preference) and the vegan package, plus various packages
in the Environmetrics Task View:

http://cran.r-project.org/web/views/Environmetrics.html

would be my choice for what you describe. A further advantage is that
new analysis techniques are being written in, or are already available
in (or can be easily ported to) R.

Of course, R has a fairly steep learning curve. I'd advise you to try to
find an introductory course near to you so you can learn to use R. Not
because you can't do it on your own (I did as a PhD student 8 years
ago), but because unless you have the time to set aside to go through
some of the introductory materials and manuals, you're likely to find R
frustrating if you jump straight in and try to do some analysis for a
job that needs to be done. At least on a course, you are there to
learn...

One of the advantages of Canoco over competing packages (IMHO, and also
based on experiences from several years ago - things may have changed in
the intervening period) was the availability of comprehensive support
for restricted permutations that allow complex, designed experiments to
be analysed via appropriate permutations tests; random shuffling is
often not suitable due to spatial/temporal autocorrelation, or
blocks/split plots etc.

This latter point applies to vegan as well of course, but we are getting
closer to the permutation functionality available in Canoco.


-----------------------------------------------------------------------
pc-ord is very easy to learn (especially if you get the book _Analysis of 
Ecological Communities_ by Bruce McCune, James B. Grace, and Dean L. Urban.. 
does an excellent job of describing the many types of analyses, the pros & 
cons, and how to do them).  you won't need to learn a whole new language as 
with R.  and it is much less expensive than CANOCO.  however,  CANOCO (and I 
think R??  not sure) give you more flexibility.  pc-ord is more of a "set up 
the matrices and push the button" kind of a program... CANOCO give much more 
user control.  for example, with CANOCO you can do all kinds of interesting 
step-wise CCA analyses, but with pc-ord you can only do one type of CCA, with 
little user control.  on the other hand, CANOCO does not do NMS (at least not 
the program i used a few years ago), which is the analyses that McCune & Grace 
push as most robust (some do not agree).

if i were just starting out, i would go with PC-ord; then pursue R or CANOCO 
later if I really got into it.

Unless you are doing marine community analysis, in which case I think most 
folks use PRIMER... at least they used to.

-----------------------------------------------------------------------
I haven't used either of the programs you listed, but Primer-e is a 
popular community ecology software package (fairly inexpensive too).

http://www.primer-e.com/

-----------------------------------------------------------------------
I've used PC-ORD for years and like it a lot for community analyses.

-----------------------------------------------------------------------
Let me just start by saying that I'm not very familiar with any of the software 
packages out there. And this might be a diversion from your main question.... 
but we had a need for NMDS recently, and I used Ginkgo. It is a slick, 
Java-based program that does many multivariate tricks 
(http://biodiver.bio.ub.es/ginkgo/Ginkgo.htm). It seems fast and might have 
some neat tricks up its sleeve.

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

I think you'll find that R is probably the most efficient route to go, even 
with the learning curve involved.  I still work outside of R to organize and 
edit data (e.g. SAS and textfiles), but once you have your datasets organized, 
running analyses are extremely easy in R and there are excellent support 
networks for getting advice/instruction on running the different packages.  The 
main one used for ordination/mulitvariate analyses is the package VEGAN.  And 
best of all, R is free!

-----------------------------------------------------------------------
I have used PC-ORD for years, and also have used PRIMER, which many folks like 
for these types of analyses (I'm assuming you are after things like ordination 
and clustering and other "distance-based" approaches to comparing community 
composition between samples). In many respects, they do the same things, 
although PC-ORD does more. I have not used CANOCO, but PC-ORD does most of the 
same things in a much more streamlined fashion. The book written by McCune and 
Jim Grace that they sell along with PC-ORD is also highly recommended as an 
intro to all of this stuff.
 
I don't use R currently, so I can't speak to it.
 
The interface in PC-ORD is a bit old-school, so you need to get used to using 
LOTUS-formatted files, but once you get past that the program works quite 
simply and it works well in the Windows environment. PRIMER works well with 
regular old XCEL files without hiccups, which can be nice, but it does less 
stuff.
 
I tend to use PC-ORD a lot more than PRIMER partly because the underlying 
analyses are derived from a variety of published sources and are more 
transparent, whereas the developers of PRIMER wrote and published new analogous 
versions of many of the analyses (for example, PC-ORD uses multiresponse 
permutation procedure MRPP and PRIMER uses ANOSIM...both are ways of testing 
whether groups of samples have significantly different community structure, but 
ANOSIM is specific to PRIMER whereas MRPP is broadly used in many stats 
packages and books).
 
I do a lot of this kind of work (mostly with bacterial communities) so if you 
have other questions about heading down this road I'd be happy to answer them.

-----------------------------------------------------------------------
I haven't used either of the programs you listed, but Primer-e is a 
popular community ecology software package (fairly inexpensive too).

http://www.primer-e.com/

-----------------------------------------------------------------------
If you haven't used either Pc-ORd or Canoco I would go with R - it does
everything those 2 packages can do and much more that neither can do, plus it
is free. Yes the learning curve is a little steeper - but there are plenty of
guides that can help you through it. Check out package "Vegan" for ordination.
That is what got me into R - I couldn't afford either PC-ord or Canoco, as a
grad student or post doc and wanted to do RDA/CCA and more.

-----------------------------------------------------------------------
R has a very steep learning curve but if you have time I'd suggest learning R 
(even though I've never used it for community analysis) just because it is 
powerful and free.  I've never seen Brodgar in action but given that it's an 
extension for R, it will matter greatly how good their documentation is. 

You may be able to find free packages for R if you're willing to do a little 
searching.  Documentation on free packages varies from almost nonexistent to 
really good.

PC-Ord (at least the version I have) is pretty picky about how you format your 
data set but once you get your data into it, it's really easy to use and has 
pretty good documentation.

I think your strategy should depend on how much time you have to devote to a 
steep learning curve.

-----------------------------------------------------------------------
I really enjoy using PC-ORD.  The only negative point that I have with the 
software is getting the data into the correct format to use it.  I have used R 
(although very limited) and for ease of use I would recommend PC-ORD.  If you 
are using direct gradient analysis (known environmental gradients) the software 
is very useful.  Also, if you have no direct gradient then you can use NMS 
which permutes the data and substitutes artificial data to give the best fits. 
 
I am running several analyses in which I have direct gradients and would highly 
recommend PC-ORD.  Another type of software that is available for non-normal 
data is HYPERNICHE (designed by McCune spp?? also) which deals with 
nonparametric data.




      

Reply via email to