Saturday, July 6, 2013

ENMTools 1.4.3

While trying to iron out the weirdness of Perl with Mac line endings in .csv files (unsuccessfully), I added some bits of code that seem to have caused the model selection functions in ENMTools to stop working on some input files.  Here's a fixed version.

13 comments:

  1. This comment has been removed by the author.

    ReplyDelete
  2. Hi Dan,

    I am trying to compare AICc scores from ENMtools to those I get from a script in R. In the same .csv for ENMtools, I specified the pathway to the files that I had created running Maxent in R as well as those models I created with Java Script. The models I created with Java work fine, and I get AICc scores for them. But the models I created in R do not get scores in ENMtools. Apparently it can find the .csv and .ascii files, but says "Opening lambdas file...Found probability of !"

    The new model_selection.csv file that is created only shows the parameters used, giving zeros for the sample size and log likelihoods.

    Any idea what I am doing wrong?

    Thanks for your time,

    Peter Galante

    ReplyDelete
  3. It's hard to know exactly why from this, but it sounds like it's not finding data at the places it thinks your points should be in the .asc file. How big is your data? Could you send me enough to run one model comparison so I can troubleshoot it?

    ReplyDelete
  4. Sure. I can send you the occurrence files, and the .lambdas and .ascii files from models made with the same features by each method.

    ReplyDelete
  5. Hi Dan,

    thanks for the continuous support and development. I've been trying to clean a csv with the no dups tool. The tool is working fine with the exact function, but when using the grid function it creates an empty csv (only the headers are copied). Obviously, there is no error message, since it thinks it created a good file.
    I am working on a W7 64 machine, with ENM 1.4.3 (the updated version), Perl64 5.16.3.



    ReplyDelete
    Replies
    1. As a follow up too this issue, I've ran the tool several times, with the same input data table and asc grid file, always with the same outcome (empty table) but eventually the tool wrote the file. I just plotted the original, resulting and asc files in a GIS and it looks like it worked. So, I do not have an answer to why this happened but my suggestion in case anyone experiences the same issue is to keep trying.

      Delete
  6. Hi Dan,
    thanks for ENMTool! I am using it to build a null model to compare it with my models. I have 31 replicates, so I used resample from raster, with 31 records and 100 replicates. I tried 1.4.3 and 1.4.2 versions in Windows, but the program run forever and then make a CVS file with row headers but no data, so I tried 10 records and 10 replicates (in order to test) and got the same results. Finally I got version 1.3 and it work; with 31 records and 200 replicates (and quite fast). So: 1. why newer versions don't work, 2. why so different in speed (this is suspicious, which one would be right for my volume of data [31 * 200 and 4 gb ram])
    3. Is there any bugs or issues about resample from raster with version 1.3, or the output should be valid for Null models (using Raes N, ter Steege H (2007) approach)?

    Thanks!

    Robert

    ReplyDelete
  7. Hi Robert, can you send me your .asc file or is it too big? It's hard to diagnose these problems without seeing the data.

    As for question 3, there aren't any major issues. The earlier versions of ENMTools sampled from rasters without replacement, which was slightly off base for the Raes and ter Steege test, but this was unlikely to have a noticeable effect except when there were very few grid cells in the study area. The newest version of ENMTools fixes it so that you can choose to sample with/without replacement, but clearly that isn't working for you. Let me take a look at your data and we'll work it out.

    Fair warning, though - I'm in Guam for fieldwork at the moment, so my ability to work on this stuff is currently somewhat limited. I'll get to it as soon as I can, though!

    ReplyDelete
  8. Hi Dan, thank you very much for the reply, and sorry to get back to you so late. Where I send you the asc file?
    The file is from worldClim dataset, but cut to my study area, not too big at all.
    I appreciate your help, but there is no much hurry as I used the 1.3 version, but it would be good to solve that problems and to use the newer version in the future models.
    Kind regards!!

    Robert Petitpas

    ReplyDelete
  9. dan.l.warren@gmail.com

    I'm in the field right now, so replies will be a bit slow.

    ReplyDelete
  10. Maybe I'm nuts or am missing something... but when I uncompress the folder there is no executable file to run. Is there another way to get this?

    ReplyDelete
  11. Unfortunately new versions of ENMTools will only be distributed as Perl scripts. The reason is discussed here:

    http://enmtools.blogspot.com.au/2012/11/testing-new-version-of-enmtools.html

    ReplyDelete
  12. Nevermind, I'm doing this through Perl.

    ReplyDelete