TWINSPAN in R

41 Replies

TWINSPAN is perhaps one of the most popular clustering methods (at least among vegetation ecologists), which is not implemented in R. R-sig-eco forum has several posts (mostly from Jari Oksanen and Dave Roberts) on the topic of TWINSPAN in R, where they described difficulties with importing the original TWINSPAN code (written in FORTRAN) into R. Seems that both Jari and Dave spent considerable effort trying to implement the method into R, but seems like there is some problem which is not easy to crack.

From my experience (and I guess also from experience of many others, not only vegetation ecologists), TWINSPAN does sometimes give rather nice and ecologically meaningful results, since it is based on cutting the data along the main compositional gradients. There is yet another divisive method, somewhat analogous to TWINSPAN, called DIANA (DIvisive ANAlysis clustering), which was proposed by Macnaughton-Smith et al. (1964), described in detail by Kaufman & Rousseeuw (1990), and made available in series of Fortran written programs with poetic names like AGNES, CLARA, DAISY, DIANA, or FANNY (later implemented into R package cluster). But it seems to me that this method for some reason never gained so much popularity like TWINSPAN did, mostly perhaps due to its sensitivity to outlying samples (which will be separated first into one-item clusters). This is why I think it still would be nice to have TWINSPAN handy in R, also because of its modified version, which we introduced couple years ago (Roleček et al. 2009) and which up to now is available only in JUICE, program for editing and analysis of vegetation data developed by Lubomír Tichý (Tichý 2002). Modified TWINSPAN is basically a sequence of divisions calculated by standard TWINSPAN, each time applied on the most heterogeneous group – here is again a similarity with DIANA, which also in each step divides the group which is the most compositionally heterogeneous.

Recently, I got a simple idea, how to make TWINSPAN work in R, and how to extend this implementation also for modified TWINSPAN. The way I used is somewhat clumsy, but it seems to work. I used the twinspan.exe file, which is an executable program based on original FORTRAN library written by Mark O. Hill (author of the algorithm and the original FORTRAN code) and compiled in 2003 by Stephan M. Hennekens into a stand-alone program for use in MEGATAB, software which before was used together with TURBOVEG for editing and classification of vegetation data (Hennekens & Schaminée 2001, Schaminée & Hennekens 2001). I created R package twinspanR, which includes twinspan.exe, and added bunch of supporting functions to maintain import and export of data to twinspan.exe and back to R. So basically the TWINSPAN is calculated by the original Hill’s algorithm, and R functions in twinspanR package are for handling the whole thing conveniently in R (see the notes below for more technical details how the library exactly works).

For details how to install the library from GitHub see the Readme.md file. Some more examples how to use twinspan package will be (hopefully) made soon available on my website for analysis of community ecology data in R.

The following is an example code, using example dataset danube with Ellenberg’s meadow dataset (used also as an example in the first publication of TWINSPAN and DCA by Hill 1979 and Hill & Gauch 1980, respectively). In this example I used modified TWINSPAN (Roleček et al. 2009) with division into 4 groups and heterogeneity of clusters measured by Bray-Curtis dissimilarity measure. I projected the results into DCA ordination diagram, along to the original tabular classification made manually by Ellenberg (from Mueller-Dombois & Ellenberg 1974):

## Modified TWINSPAN on traditional Ellenberg's Danube meadow dataset,
## projected on DCA and compared with original classification into
## three vegetation types made by tabular sorting:
library (twinspanR)
library (vegan)
data (danube)
res <- twinspan (danube$spe, modif = TRUE, clusters = 4)
k <- cut (res)
 
dca <- decorana (danube$spe)
par (mfrow = c(1,2))
ordiplot (dca, type = 'n', display = 'si', main = 'Modified TWINSPAN')
points (dca, col = k)
for (i in c(1,2,4)) ordihull (dca, groups = k, show.group = i, col = i,
 draw = 'polygon', label = TRUE)
ordiplot (dca, type = 'n', display = 'si', main = 'Original assignment\n (Ellenberg 1954)')
points (dca, col = danube$env$veg.type)
for (i in c(1:3)) ordihull (dca, groups = danube$env$veg.type,
 show.group = unique (danube$env$veg.type)[i], col = i,
 draw = 'polygon', label = TRUE)

Some technical details how the twinspanR package works

Some further details how it works. The executable file, twinspan.exe, is stored in \exec subdirectory of the library. There is also tw.bat file, which launches twinspan.exe and feeds it with data from R (I have shamelessly stolen this idea from the way how Lubomír Tichý executes TWINSPAN in JUICE, and it is also similar to the way how Tom August implemented another Hill’s FORTRAN library, FRESCALO (Hill 2011), into R as a function frescalo in package sparta). I take compositional data in R, transform them to required *.cc! format (luckily there is a function write.CEP in rioja package, written by Steve Juggins) and save them to \exec subdirectory (where also twinspan.exe is located). Then, I create file tw.dat with input parameters for twinspan.exe (using function create.tw.dat), and use the shell function in R to launch the tw.bat file. All calculations are done using original twinspan.exe, R just reads its output from tw.PUN file. It’s in a no way an elegant approach, but works just fine.

The twinspan.exe file used for calculation of TWINSPAN in R is taken from the distribution of JUICE program (Tichý 2002). It is the original FORTRAN code written by Mark O. Hill, compiled (cca in 2002) by Stephan M. Hennekens into twinspan.exe for use in his program MEGATAB (which was used together with Turboveg for analysis of community data). The version of twinspan.exe used here implements the changes by Petr Šmilauer, related mainly to problems with algorithm convergence, which cause the results being dependent on the order of samples in the input data table. Note that this algorithm is slightly different from TWINSPAN implemented in WinTWINS software (Hill & Šmilauer 2005), which implements also other modifications by Birks and ter Braak. In fact, there are at least four different versions of TWINSPAN recently used for analysis, and in certain circumstances they differ in the results how they classify the samples (when I have time, I will elaborate this topic further).

Unfortunately, the presence of twinspan.exe file in the twinspanR library is a problem for its portability – seems like CRAN doesn’t allow packages with executables inside (for understandable security reasons), and R-Forge allows it to be uploaded, but fails to build it. For now, the channel for distribution of this package is GitHub, and it will perhaps remain there until some other solution will show up. Simply, it’s just a quick and dirty way how to get TWINSPAN functionality in R without too much hassle, before somebody manages to write fully functional implementation of TWINSPAN in R.

References

Hennekens S.M. & Schaminée J.H.J. (2001): TURBOVEG, a comprehensive data base management system for vegetation data. Journal of Vegetation Science, 12: 589-591.
Hill M.O. (1979): TWINSPAN: A FORTRAN Program for Arranging Multivariate Data in an Ordered Two-way Table by Classification of the Individuals and Attributes. Cornell University, Ithaca, NY.
Hill M.O. & Šmilauer P. (2005): TWINSPAN for Windows version 2.3. Centre for Ecology and Hydrology & University of South Bohemia, Huntingdon & České Budějovice.
Hill M.O. & Gauch H.G. (1980): Detrended correspondence analysis: An improved ordination technique. Vegetatio, 42: 47-58.
Kaufman L. & Rousseeuw P. J. (2009): Finding groups in data: an introduction to cluster analysis (Vol. 344). John Wiley & Sons.
Macnaughton-Smith P., Williams W. T., Dale M. B. & Mockett L. G. (1964): Dissimilarity analysis: a new technique of hierarchical sub-division. Nature, 202: 1034-1035.
Mueller-Dombois D. & Ellenberg H. (1974): Aims and Methods of Vegetation Ecology. John Wiley & Sons, New York, Chichester, Brisbane, Toronto.
Roleček J., Tichý L., Zelený D. & Chytrý M. (2009): Modified TWINSPAN classification in which the hierarchy respects cluster heterogeneity. Journal of Vegetation Science, 20:596-602.
Schaminée J.H.J & Hennekens S.M. (2001): TURBOVEG, MEGATAB und SYNBIOSYS: neue Entwicklungen in der Pflanzensoziologie. Ber. Reinhold-Tüxen Ges., 13: 21-34.
Tichý L. (2002): JUICE, software for vegetation classification. Journal of Vegetation Science, 13: 451-453.

41 thoughts on “TWINSPAN in R”

Attila Lengyel October 13, 2015 at 11:29 am

Hi David,
it is nice to see Twinspan finally implemented in R! Once I also tried to write a script for it, however, I was unable to follow all the ‘tricks’ hidden in the original Twinspan code but left out from the descriptions, thus I gave up with it. That’s why it’s great that your programme uses the ‘original’ (or at least a trusted) version.
Congratulations for your especially useful site!
Cheers,
Attila

Reply ↓
1. David Zelený October 13, 2015 at 11:43 am
  
  Hi Attila,
  Thanks for a nice comment! I was working on TWINSPAN in R last Christmas, because I just suddently realized the way how it can be done in R and just wanted to try it – and looks like it works. The package is far from ready, but is somewhat functional – there is still couple of things needed to be finished, like plotting the dendrogram, or printing the indicator values for individual splits. Hope to finish these once. I will be happy, if you use it, if you give me a feedback – especially about things which are not working in a way advertised in the package.
  
  Have a nice day!
  
  David
  
  Reply ↓
Sapphire McMullan-Fisher January 4, 2016 at 10:33 pm

I can’t seem to fine your twinspanR library in CRAN – did I miss a link or is it not yet available?

Best wishes
Sapphire

Reply ↓
Sapphire McMullan-Fisher January 4, 2016 at 11:03 pm

Apologies for previous question. I am new to R and didn’t realise you could get other packages from places other than CRAN – managed to get the package out of github.

Is there a twinspan users or chat group?

Best wishes
Sapphire

Reply ↓
1. David Zelený January 5, 2016 at 2:41 pm
  
  Hi Sapphire,
  
  you are right, the package is only on GitHub, and this is perhaps not going to change any time soon – the problem is that the twinspanR library can be used only under Windows (due to it’s internal dependencies, since the TWINSPAN algorithm is calculated externally using twinspan.exe program), so it cannot be uploaded into CRAN (which requires multiplatform usability of packages). Indeed, this is only one reason – the package is in the developmental stage, and was a kind of experiment – hope to finish in future when I have more time, if I see it can be useful. CRAN is not a good place for packages under development, while GitHub is.
  
  Cheers!
  
  David
  
  Reply ↓
  1. Sapphire McMullan-Fisher January 26, 2016 at 6:22 am
    
    Thanks for you reply. It may have also answered the question I had when I ran twinspan as only half of it worked with the test data and I am using a mac 🙁
    
    Are you planning on making it mac friendly at some point? I will see if I can find someone with R on a windows machine.
    
    Thanks
    Sapphire
    
    Reply ↓
David Zelený January 26, 2016 at 7:05 am

Hi Sapphire,

To write a Twinspan package which can be used across platforms would be great, but currently I don’t plan to do it – it’s different story than my simple solution (to wrap twinspan.exe file and send data into and back). The most recent attempt to rewrite twinspan into R I know is from Dave Roberts, but his solution differs from “genuine” TWINSPAN of Mark Hill, so the results will be more or less different. I also heard somewhere that Jari Oksanen tried to port the original Fortran library into R, but didn’t succeed for some reason, but may be if he has motivation, he may give it a try again. That’s it, sorry for that 🙂

David

Reply ↓
kieran February 24, 2016 at 5:31 pm

Hi David,

Very useful to have an R compatible TWINSPAN! Congratulations and thank you for your efforts.
I’m interesting in reviewing the hierarchical organization of my data (via twinspan), could you tell me if it’s possible to plot a dendrogram directly from the output? And if not, do you have any tips as to how I might (most efficiently) extract and organize the relevant information from the output provided?

Thanks, Kieran.

Reply ↓
1. David Zelený February 25, 2016 at 11:48 pm
  
  Hi Kieran,
  
  Unfortunately there is no plotting function in the twinspanR package so far, I haven’t found a handy way how to make it. So far the best guess is to extract the hierarchy from the output and draw it somewhere externally. If you use standard Twinspan, just print the two-way sorted table, and at the bottom of the table you will find the 0-1 sequence determining the hierarchy, e.g.:
  tw.res <- twinspan (danube$spe) print (tw.res, what = 'table')
  
  The same 0-1 sequence is also in twinspan output in the variable “class”.
  
  If you use modified twinspan, the table output is more complex (there is table for each division), but the “class” output should help:
  tw.res.modif <- twinspan (danube$spe, modif = T) tw.res.modif
  order plot.no class 1 1 1 *1111 2 2 2 *1111 3 3 3 *1111 4 4 4 *1111 5 5 5 *0000 6 6 6 *0011 7 7 7 *0011 ...
  
  I should make the dendrogram output, just so far I didn’t found a clever way how to do it…
  
  Hope this helps
  
  David
  
  Reply ↓
kieran February 26, 2016 at 2:50 pm

Hi David,
Thanks for your comments. So it seems exporting and manual construction is the way to go if I want to create the dendrogram itself. Although I guess I can actually review the hierarchical group structure quite easily (in a piecemeal fashion) by interrogating the binary matrix (with site labels), which would serve my current analytical desires just fine….
Anyways, thanks again for sorting a quick and easy TWINSPAN option(s) for R and for your subsequent help!

All the best,
Kieran.

Reply ↓
Noah Dell March 3, 2016 at 12:00 am

David,

I was wondering if you might know (or else be able to find) what the values for the convergence criteria were set as in the Hill and Šimlauer (2005) version of TWINSPAN. Are the convergence criteria in that version of TWINSPAN and that implemented in twinspanR similar to those suggested by Oksanen and Minchin (1997)?

Thanks,

Noah

Reply ↓
1. David Zelený March 8, 2016 at 12:11 am
  
  Hi Noah,
  
  My guess is that the convergence criteria are the same as in Hill & Šmilauer 2005. The reason for my guess is that twinspanR is using twinspan.exe library, which is currently used in JUICE software, and as I know, Petr Šmilauer couple of years ago was fixing this file, since it gave unstable classification results. But there are some differences between twinspan.exe and Hill & Šmilauer’s Windows program, since the later implements also other tweaks by ter Braak and Birks (I guess), which are not implemented in twinspan.exe (this has different evolution – it comes from original Hill’s fortan-based twinspan, and was originally maintained by Stephan Hennekens for his Megatab software).
  I did some testing, though not really comprehensively, but seems like each version is giving slightly different results (in non standard parametric settings). Problem is that only the original Hill’s algorithm is somewhat properly described (in his report from 1979, which is unfortunately hard to get), the later changes are actually not properly described anywhere.
  
  Not sure if this helps…
  
  David
  
  Reply ↓
  1. Noah Dell March 8, 2016 at 2:04 am
    
    David,
    
    It does help quite a bit. I am examining different methods for clustering/classification of vegetation data and was glad to find your beta version of twinspanR. I have since been looking into what is different between versions of TWINSPAN.
    
    Thanks again,
    
    Noah
    
    Reply ↓
Ivy April 8, 2016 at 2:58 pm

Hello David,
I’m new in using R. I checked my repositories an I do not have Github. What do I do?

Reply ↓
1. David Zelený April 8, 2016 at 3:34 pm
  
  Hi Ivy,
  Don’t worry about Github – you don’t need it on your computer, it’s online repository where the twinspanR library is stored. To install the twinspanR library, first install library devtools, and then twinspanR:
  install.packages ('devtools') devtools::install_github("zdealveindy/twinspanR") library (twinspanR)
  After this, you should be able to use TWINSPAN in R.
  Cheers!
  David
  
  Reply ↓
Noah Dell May 17, 2016 at 10:39 pm

David,

I have encountered an error with twinspanR. Sometimes when I go to use it, I get an error that reads:

“Error in groups01[,1] = cut(tw.temp[[1]],level=1)-1: number of items to replace is not a multiple of replacement length”

Is this something you have encountered before? I have looked over the source code and cannot exactly figure out what this means. My guess is that there is something happening between the creation of the empty matrix group01 and the empty list tw.temp and the the reference to tw.temp[[1]], and I presume it is happening in the call to twinspan0.

Thanks in advance,

Noah

Reply ↓
1. David Zelený May 17, 2016 at 11:05 pm
  
  Hi Noah,
  
  Wouldn’t you post (or send) some small dataset which is causing this error? Just to be able to reproduce… I will have a look what’s going on…
  
  David
  
  Reply ↓
  1. Noah Dell May 17, 2016 at 11:56 pm
    
    David,
    
    I have sent, in an email, an example data set that has been acting up on me, though it is not the only one.
    
    Thanks again,
    
    Noah
    
    Reply ↓
    1. David Zelený May 18, 2016 at 7:06 am
      
      Hi Noah,
      
      I think I found the bug, thanks for sending the data. The problem is actually with row names in your community matrix – the rownames should not contain the white spaces, this will cause trouble in ‘cut.tw’ function. The easiest way to fix is to avoid the spaces at all. I modified the twinspanR library (function twinspan) – I released the twinspanR version v0.16 into GitHub which – in case that the rownames contain spaces will convert it into syntactically valid names using ‘make.names’ function (replace spaces with dots).
      
      Can you check whether it works with the v0.16 version and let me know? When I tried, looks ok, no error.
      
      Cheers
      David
      
      Reply ↓
Noah Dell May 18, 2016 at 3:08 pm

David,

That works perfectly. I can’t believe I had not thought of that. That makes sense why the program seemed to “think” that it had two different numbers of elements at different points. I even had made modifications to print the output from certain functions and was very confused because it seemed to be working internally.

Thanks again,
Noah

Reply ↓
1. David Zelený May 18, 2016 at 3:13 pm
  
  Great, happy that it works. Thanks for feedback, it’s great to know that somebody is actually really using that 🙂 This library is just my sentiment, once during Christmas I just by the way found a way how to put twinspan.exe into R folder and tailor all those algorithms to feed it by data and take data out, and cannot get rid of the excitements that it may be actually “twinspan in R” until it was finished. What bugs me most at this moment is that there is not good plot function for the dendrogram, I want to add, but it’s a bit complex…
  Cheers!
  David
  
  Reply ↓
Hugo December 19, 2016 at 5:37 am

Frist of all, thank you so much! Hope I can use the twinspanR in mac someday.
Have a nice day~!

Reply ↓
Jolien D May 24, 2017 at 9:18 pm

Hi

Thank you for creating this package, but I don’t seem to be able to install it. I’ve tried with the directions you gave with devtools and I’ve even tried downloading the zip and then installing it from the local file, but it always seem to uninstall itself for some reason, when trying to install I always get this error:

Downloading GitHub repo zdealveindy/twinspanR@master
from URL https://api.github.com/repos/zdealveindy/twinspanR/zipball/master
Installing twinspanR
Installing 1 package: vegan
Installing package into ‘C:/Users/jolientje/Documents/R/win-library/3.2’
(as ‘lib’ is unspecified)
Warning: package ‘vegan’ is in use and will not be installed
“C:/PROGRA~1/R/R-32~1.2/bin/i386/R” –no-site-file –no-environ –no-save –no-restore
–quiet CMD INSTALL
“C:/Users/jolientje/AppData/Local/Temp/Rtmpusp9vO/devtools1f947e2b7e4b/zdealveindy-twinspanR-63e085f”
–library=”C:/Users/jolientje/Documents/R/win-library/3.2″ –install-tests

* installing *source* package ‘twinspanR’ …
** R
** data
** exec
** preparing package for lazy loading
Warning: package ‘lattice’ was built under R version 3.2.3
Warning: package ‘rioja’ was built under R version 3.2.5
Warning: package ‘betapart’ was built under R version 3.2.5
Error in loadNamespace(j <- i[[1L]], c(lib.loc, .libPaths()), versionCheck = vI[[j]]) :
there is no package called 'geometry'
Error : package 'betapart' could not be loaded
ERROR: lazy loading failed for package 'twinspanR'
* removing 'C:/Users/jolientje/Documents/R/win-library/3.2/twinspanR'
Error: Command failed (1)

Am I doing something wrong or are my settings wrong?
Do you have any idea how I can solve this?

Reply ↓
1. david Post authorOctober 15, 2017 at 4:30 am
  
  Hi Jolien,
  
  Sorry, I overlooked this comment, now I see you posted it already in May, sorry for that! I guess the reply is already not relevant for you, but could be for others running into the same problem.
  
  From the error message you posted it seems like a problem with R package dependencies, not really twinspanR package. Looks like ‘betapart’ package cannot load the dependency ‘geometry’, for reason I cannot see. Could you try to install ‘betapart’ package first, i.e. type
  
  install.packages (‘betapart’)
  
  and then run the twinspanR installation again? If you observe similar error messages, just install the problematic library manually as the betapart library above (e.g. install manually the library ‘geometry’, which was causing the problem in your case).
  
  David
  
  Reply ↓
Arayaslassie Abebe April 12, 2019 at 8:34 pm

Error in curl::curl_fetch_memory(url, handle = h) :
Failed to connect to api.github.com port 443: Timed out
> this message comes what shall I do?

Reply ↓
1. David Zelený Post authorApril 12, 2019 at 10:31 pm
  
  Hi, can you provide R code to show at which situation this happens?
  
  Reply ↓
Wangya Han May 10, 2019 at 9:46 am

Hi, David,
There are some wrong when I install package twinspanR. Your help is badly needed! Thank you!
R code as following:
> devtools::install_github(“zdealveindy/twinspanR”)
Downloading GitHub repo zdealveindy/twinspanR@master
WARNING: Rtools is required to build R packages, but is not currently installed.

Please download and install Rtools 3.5 from http://cran.r-project.org/bin/windows/Rtools/.

√ checking for file ‘C:\Users\680S1S2\AppData\Local\Temp\RtmpkVwyxW\remotes55343b63a7a\zdealveindy-twinspanR-63e085f/DESCRIPTION’ (434ms)

– preparing ‘twinspanR’:
checking DESCRIPTION meta-information …

√ checking DESCRIPTION meta-information

– checking for LF line-endings in source and make files and shell scripts

– checking for empty or unneeded directories

– looking to see if a ‘data/datalist’ file should be added

– building ‘twinspanR_0.16.tar.gz’

* installing *source* package ‘twinspanR’ …
** using staged installation
** R
** data
** exec
** byte-compile and prepare package for lazy loading
错误: object ‘write.CEP’ is not exported by ‘namespace:rioja’
停止执行
ERROR: lazy loading failed for package ‘twinspanR’
* removing ‘C:/Program Files/R/R-3.6.0/library/twinspanR’
Error in i.p(…) :
(由警告转换成)installation of package ‘C:/Users/680S1S2/AppData/Local/Temp/RtmpkVwyxW/file5534204067f7/twinspanR_0.16.tar.gz’ had non-zero exit status

Reply ↓
1. David Zelený Post authorJune 12, 2019 at 9:26 am
  
  Hi, sorry for late reply. This problem happens because the package rioja no more contains the function write.CEP, needed by twinspanR package. The workaround is to install the riojaExtra package first, directly from GitHub repository, using the command
  devtools::install_github(“nsj3/riojaExtra”)
  Then, install the latest version of twinspanR, it should work (if not, let me know). Details here: https://github.com/zdealveindy/twinspanR
  D
  
  Reply ↓
Naghmeh June 12, 2019 at 2:16 am

Dear Dr. Zeleny
Thanks for creating this package. But I can’t use twinspanR library .

I used this steps for install:

install.packages (‘devtools’)
devtools::install_github(“zdealveindy/twinspanR”)
library (twinspanR)

Error in library(twinspanR) : there is no package called ‘twinspanR’

then I read commnts in this page. You sugessted that install ” betapart” and ” geometry” but after install these. I fail to use this package. I wondering if you could take valuble comments to slove the problem?

Sincerely

Reply ↓
1. David Zelený Post authorJune 12, 2019 at 9:29 am
  
  Hi Naghmeh,
  Installing betapart and geometry was an old problem, I think this is not an issue any more. But current problem is that the riojaExtra library needs to be installed from GitHub directly. Please, follow instructions in https://github.com/zdealveindy/twinspanR – first install riojaExtra, and after you succeed, install twinspanR.
  Hope it works,
  David
  
  Reply ↓
Bukombe June 22, 2019 at 9:58 am

Hi
I am still lerning r and want to use TWINSPAN, After following some dicusion I realized I should install TwinspanR using the following
install.packages (‘devtools’) # if you don’t have it installed yet
devtools::install_github(“nsj3/riojaExtra”)
devtools::install_github(“zdealveindy/twinspanR”)

After doing this I still get trouble as I still get an error message below
These packages have more recent versions available.
Which would you like to update?

1: vegan (2.5-4 -> 2.5-5) [CRAN]

Enter one or more numbers separated by spaces, or an empty line to cancel

Please any one to help

Reply ↓
1. David Zelený Post authorJune 24, 2019 at 11:42 am
  
  Hi Bukombe, this is just a warning that newer version of vegan, one of the packages required by twinspanR, has a newer version available. If you run into this, simply press 1 to update. Alternatively, you can in R studio update the vegan package manually. Remember that to update vegan, you need to restart R to make sure that it is not loaded (otherwise it will break the package structure due to the dynamic library loaded). If you are not sure how to do it, then in RStudio menu, go Session > Restart R. Then, use command line to write
  remove.packages(“vegan”)
  install.packages (“vegan”)
  This should uninstall the old vegan version and install the new one.
  Hope it helps,
  David
  
  Reply ↓
Linda August 20, 2019 at 6:34 pm

Hi David,
I’m sorry to bother you. I have a problem similar to Wangya Han.
There are some wrong when I install package riojaExtra. The description of this problem is as follows.

devtools::install_github(“nsj3/riojaExtra”)
Downloading GitHub repo nsj3/riojaExtra@master
Your system is ready to build packages!

√ checking for file ‘C:\Users\两两\AppData\Local\Temp\RtmpAtf7D8\remotes64034d65b24\nsj3-riojaExtra-506dbe1/DESCRIPTION’ (364ms)

– preparing ‘riojaExtra’:
checking DESCRIPTION meta-information …

checking DESCRIPTION meta-information …

√ checking DESCRIPTION meta-information
– cleaning src

– installing the package to process help pages

———————————–

– installing *source* package ‘riojaExtra’ …
** using staged installation
Warning in file(file, if (append) “a” else “w”) :
鏃犳硶鎵撳紑鏂囦欢’C:/Users/A=A=/AppData/Local/Temp/RtmpIlvJnp/Rinst43f8639c3281/00LOCK-riojaExtra/00new/riojaExtra/DESCRIPTION’: No such file or directory
Error in file(file, if (append) “a” else “w”) : 鏃犳硶鎵撳紑閾剧粨
ERROR: installing package DESCRIPTION failed for package ‘riojaExtra’
– removing ‘C:/Users/两两/AppData/Local/Temp/RtmpIlvJnp/Rinst43f8639c3281/riojaExtra’
———————————–
ERROR: package installation failed

错误: Failed to install ‘riojaExtra’ from GitHub:
System command error, exit status: 1, stdout + stderr (last 10 lines):
E> ———————————–
E> * installing *source* package ‘riojaExtra’ …
E> ** using staged installation
E> Warning in file(file, if (append) “a” else “w”) :
E> 鏃犳硶鎵撳紑鏂囦欢’C:/Users/A=A=/AppData/Local/Temp/RtmpIlvJnp/Rinst43f8639c3281/00LOCK-riojaExtra/00new/riojaExtra/DESCRIPTION’: No such file or directory
E> Error in file(file, if (append) “a” else “w”) : 鏃犳硶鎵撳紑閾剧粨
E> ERROR: installing package DESCRIPTION failed for package ‘riojaExtra’
E> * removing ‘C:/Users/两两/AppData/Local/Temp/RtmpIlvJnp/Rinst43f8639c3281/riojaExtra’
E> ———————————–
E> ERROR: package installation failed

I don’t know where something went wrong. By the way, my R version is 3.6.1. Anyway, I have to thank you for your help.
Your sincerely,
Linda.

Reply ↓
1. David Zelený Post authorAugust 22, 2019 at 2:06 pm
  
  Hi Linda,
  I see you cannot install the riojaExtra package from GitHub. From the error message, it seems to me that the problem is on the side of your computer. Your user name is using Chinese characters (两两), which in a certain moment will display as “A=A=” (see the warning message starting with 鏃犳硶鎵撳紑鏂囦欢). I guess the problem may be that the R is unable to install the package into the folder which contains Chinese characters in the path. This is actually a common problem, not related only to installing packages in R, but also to using RStudio (I know what I am talking about since I am teaching R in Taiwan, where many students face the same issue). The solution: create a new user on your computer, which will not contain the Chinese characters (e.g. Linda), switch to this user and install the R and related packages from there. I guess it should work.
  Cheers,
  David
  
  Reply ↓
  1. David Zelený Post authorAugust 22, 2019 at 2:08 pm
    
    … also this link may help: https://github.com/wush978/DataScienceAndR/wiki/Windows%E4%B8%AD%E6%96%87%E4%BD%BF%E7%94%A8%E8%80%85%E8%88%87Rstudio%E7%9A%84%E7%92%B0%E5%A2%83%E8%AE%8A%E6%95%B8%E8%AA%BF%E6%A0%A1
    
    Reply ↓
Yizhen Liu December 2, 2019 at 10:49 pm

Hi David，
Thx for the wonderful package. But I have some problem to trouble you.
(1) I want to know how to export the result when I have achived twinspan, especially of “twi”, because the result of “twi” is similarly to the result of twinspan in PCORD5.0. When I use write.table to export the twi as a txt, it is messy for me to read as follow:
“x”
“1” ” @”
“2” ” ”
“3” ” This version of TWINSPAN allows you to specify WEIGHTS for samples and species at the input device ”
“4” ” Reading data matrix from device 5″
“5” ” 1331 212 tw.cc! ”
“6” ” Number of SAMPLES 212″
“7” ” Number of SPECIES 1331″
“8” ” LENGTH of raw DATA array 18012″
“9” ” SPECIES NAMES”
“10” ” 1 Aca evoT| 2 Aca senS| 3 Ace bueT| 4 Ace jiuT| 5 Ace calT| 6 Ace corT| 7 Ace davT| 8 Ace davS”
“11” ” 9 Ace eleT| 10 Ace eleS| 11 Ace fabT| 12 Ace fabS| 13 Ace flaT| 14 Ace flaS| 15 Ace lucS| 16 Ace oblS”
“12” ” 17 Ace oliT| 18 Ace oliS| 19 Ace palT| 20 Ace palS| 21 Ace sinT| 22 Ace sinS| 23 Ace truT| 24 Ace tutT”
“13” ” 25 Ace tutS| 26 Ace wilT| 27 Ac hbid| 28 Ac hjap| 29 Ac otat| 30 Act eriS| 31 Act latS| 32 Ad elav”
“14” ” 33 Ad icap| 34 Ad ifla| 35 Adi pilT| 36 Adi pilS| 37 Adi rubS| 38 Adi gliT| 39 Adi milT| 40 Adi milS”
……
“227” ” DIVISION 1 (N= 212) I.E. GROUP * ”
“228” ” Eigenvalue 0.473 at iteration 5″
“229” ” INDICATORS, together with their SIGN”
“230” ” Pt emul1(+)”
(I have two question in this section: a. what the meaning of the number after the species ,such as the 1 after Ptemul; b. why did the function slipt the species name with blank space automated)
“2800” ” 1 111111111111111 1111 1111111111111112 11111 111111″
“2801” ” 99222 1111111222 1222233333345550111166666777893299958888888045667788889267189999000045000012″
“2802” ” 2334560236789067123457891128901248980148125634569234445203822345678028018907891170190129123433567934″
“2803” ” 441 Emb rudS ——-1-3——22–122111222—22-3323—1423————————2———————-5—— 000100″
“2804” ” 409 Dr ydec ——————55—–3–4-5–33—-3——43–3—————————————-35——- 000101″
“2805” ” 85 Ant bunS –222-221122——-11132232122—1-1223—12———————————————–222—— 000110″
“2806” ” 174 Cam salS –42-33355-5–53——–1—–55——————————————————————– 000110″
(I don`t think this kind of twinspan table is regularity enough to read. And how can I achieve a twinspan table like the figure in this blog？）

(2) If I can plot a dendrogram with this twinspan result, or obtain the data of dissimilarty distance of each plots to plot a dendrogram with the twinspan result? I can`t find the dissimilarty distance data in the twinspan result after use twinspan function in TWINSPANR.

Sincerely!
Yizhen Liu

Reply ↓
1. David Zelený Post authorMarch 14, 2020 at 6:16 pm
  
  Hi Yizhen,
  Sorry for being late with this response.
  
  1) You can export the results of “twi” into the txt file directly using the combination of write.table, capture.output and print.tw function (see Example section in the help file for the function print.tw). Here, in print.tw, the argument “what” should be “what = ‘twi'”:
  
  data (danube)
  tw <- twinspan (danube$spe) write.table (file = 'table.txt', capture.output (print (tw, what = 'twi')), quote = F, row.names = F) system (shell ('notepad.exe table.txt')) The output is then as it shows in the original file: This version of TWINSPAN allows you to specify WEIGHTS for samples and species at the input device Reading data matrix from device 5 94 25 tw.cc! Number of SAMPLES 25 Number of SPECIES 94 LENGTH of raw DATA array 1601 SPECIES NAMES 1 Achi mill| 2 Ajug rept| 3 Alch vulg| 4 Alop prat| 5 Ange sylv| 6 Anth silv| 7 Arrh elat| 8 Bell pere 9 Brac pinn| 10 Briz medi| 11 Brom erec| 12 Camp glom| 13 Camp rotu| 14 Card prat| 15 Care acut| 16 Care flac 17 Care grac| 18 Care hirt| 19 Care pani| 20 Cent jace| 21 Cera caes| 22 Chen albu| 23 Chry leuc| 24 Cirs oler ...... ...... ...... 2) No dendrogram plotting function, sorry for that, I didn't really have time to add it. If you are interested, don't you want to clone the github repository and try to do that? (should not be too difficult, I think it can use similar strategy as hclust package does). Or, consider checking the new twinspan package by Jari Oksanen (https://github.com/jarioksa/twinspan), which has much better functionality (but, unfortunately, does not support modified TWINSPAN).
  
  Hope it helps,
  
  David
  
  Reply ↓
Leslie March 11, 2020 at 11:31 pm

Hi David,
I’m sorry to trouble you. There are some wrong when I install package riojaExtra. The description of this problem is as follows.

> install.packages(‘devtools’)
WARNING: Rtools is required to build R packages but is not currently installed. Please download and install the appropriate version of Rtools before proceeding:

https://cran.rstudio.com/bin/windows/Rtools/
Installing package into ‘C:/Users/sq/Documents/R/win-library/3.6’
(as ‘lib’ is unspecified)
试开URL’https://mirrors.tuna.tsinghua.edu.cn/CRAN/bin/windows/contrib/3.6/devtools_2.2.2.zip’
Content type ‘application/zip’ length 351275 bytes (343 KB)
downloaded 343 KB

package ‘devtools’ successfully unpacked and MD5 sums checked

The downloaded binary packages are in
C:\Users\sq\AppData\Local\Temp\RtmpKYjocu\downloaded_packages
> devtools::install_github(“nsj3/riojaExtra”)
WARNING: Rtools is required to build R packages, but the version of Rtools previously installed in C:/Users/sq/Documents/R/win-library/3.6/rtools40 has been deleted.

Please download and install Rtools custom from http://cran.r-project.org/bin/windows/Rtools/.
Downloading GitHub repo nsj3/riojaExtra@master
Error in utils::download.file(url, path, method = method, quiet = quiet, :
无法打开URL’https://api.github.com/repos/nsj3/riojaExtra/tarball/master’
> devtools::install_github(“zdealveindy/twinspanR”)
WARNING: Rtools is required to build R packages, but the version of Rtools previously installed in C:/Users/sq/Documents/R/win-library/3.6/rtools40 has been deleted.

Please download and install Rtools custom from http://cran.r-project.org/bin/windows/Rtools/.
Downloading GitHub repo zdealveindy/twinspanR@master
Skipping 1 packages not available: riojaExtra
Installing 9 packages: riojaExtra, betapart, magic, lpSolve, linprog, RcppProgress, fastmatch, geometry, rcdd
Installing packages into ‘C:/Users/sq/Documents/R/win-library/3.6’
(as ‘lib’ is unspecified)
错误: Failed to install ‘twinspanR’ from GitHub:
(由警告转换成)package ‘riojaExtra’ is not available (for R version 3.6.2)

My R version is 3.6.2, and I have downloaded and installed rtools40, or /rtools35 .I’ve tried downloading and reinstalling several times, but there’s always a problem.I don’t know where something went wrong. Anyway, I have to thank you for your help.
Your sincerely,
Leslie

Reply ↓
1. David Zelený Post authorMarch 14, 2020 at 6:06 pm
  
  Hi Lesli,
  Looks really like a problem with Rtools installed on your computer (it does not want to compile the riojaExtra package from GitHub). I think with R version 3.6.2 you need to use Rtool 35, not 40 (that one is only a beta version for R 4.0). Not sure how you install Rtools, but if I can suggest, do it manually – if you install some already, consider uninstalling all of them (they are usually in c:\Rtools\ folder), download the file from https://cran.r-project.org/bin/windows/Rtools/ (the Rtools 35 version) and install it. I have bad experience with installing Rtools from RStudio directly, usually something goes wrong. When installing Rtools, make sure that you add their path into the registry (check how to do it on my website here: https://anadat-r.davidzeleny.net/doku.php/en:r#install_rtools_for_windows_users_only). Then, try to install devtools (if not installed yet) and riojaExtra again. I just did on my comp (R3.6.2, Rtools 35) and it works, I managed (after updating also other packages like Rccp etc) to install twinspanR successfully.
  David
  
  Reply ↓
Jen Weaver December 14, 2021 at 4:21 am

Hi! I am trying to run twinspan from within RStudio but it causes a fatal error every time. I suspect it is because I don’t have admin privileges, and therefore the executable can’t run. Have you come across this? Do you have any suggestions? I do have an admin password I can enter but I’ve only ever done that through a GUI.

Reply ↓
1. David Zelený Post authorFebruary 13, 2022 at 3:18 pm
  
  To be honest I haven’t encountered this situation; make sure that you install the R library somewhere where you don’t need admin password, ie I install my packages in Documents folder on the computer (C:/Users/zeleny/Documents/R/win-library/4.1/”), which doesn’t need admin access. If not solving the issue, can you share the error message here?
  
  Reply ↓