[Biomod-commits] Projection Error (sum (ProbData) !=0)

[David Roberts]

Hi Bruno,

 

Yes, indeed this is the case. The FDA process produces NA outputs in the projections, so either removing the Bin, etc. projections or setting FDA=F clears up the error. Oddly, it only produces NA outputs for some species and not for others, even though the xyz for the projections points for all species is all the same (i.e. it produces projections for some species but only NA for others given the same environmental data, but non consistently, as all species have at least some valid (numerical) FDA outputs).

 

I will re-build the models and try the projections again, but this takes some time (about a day). Will also look at the enviro data to see if there’s a pattern in the NA values. Stay tuned...

 

Thanks for the help.

-          David

 

________________________

 

David Roberts

PhD Student

Dept. of Renewable Resources

University of Alberta

(780) 492-2540

 <mailto:drr3 at ualberta.ca> drr3 at ualberta.ca

 

From: Bruno Lafourcade [mailto:brunolafourcade at aol.com] 
Sent: Thursday, June 17, 2010 6:36 AM
To: drr3 at ualberta.ca
Subject: Re : [Biomod-commits] Projection Error (sum (ProbData) !=0)

 

Hum that's funny...

I see no reasons why the FDA T/F would make any difference. Perhaps the models built did something funny.
The error you get is to do with NAs somewher in the data, I am guessing it is in what the FDA produces as 
a projection. 

To check this, could you try it with FDA=T and BinRoc=F, BinKappa=F, BinTSS=F, FiltRoc=F, FiltKappa=F, FiltTSS=F 

(if no conversion of data is needed to binary or filtered there shouldn't be a problem with having NAs in it) and check what the FDA projection
looks like.

Thanks !
Bruno

 

-------
Bruno Lafourcade
Statistical tools engineer

Laboratoire d'Ecologie Alpine, bureau 308
CNRS - UMR 5553, 2233 rue de la piscine
38400 Saint Martin d'Hères
-------

 

 

-----E-mail d'origine-----
De : David Roberts <drr3 at ualberta.ca>
A : biomod-commits at r-forge.wu-wien.ac.at
Envoyé le : Mercredi, 16 Juin 2010 21:33
Sujet : Re: [Biomod-commits] Projection Error (sum (ProbData) !=0)

Update: model projections run smoothly when “FDA=F”. Perhaps I am missing an FDA setting? Using “MDA=T” instead (on a hunch) returns an unused argument error on MDA=T. For the record, the model built smoothly with FDA=T.

 

Apologies if you know this already and I missed a previous string. 

 

-          DR

 

________________________

 

David Roberts

PhD Student

Dept. of Renewable Resources

University of Alberta

(780) 492-2540

drr3 at ualberta.ca

 

From: David Roberts [mailto:drr3 at ualberta.ca] 
Sent: Wednesday, June 16, 2010 11:31 AM
To: 'biomod-commits at lists.r-forge.r-project.org'
Subject: Projection Error (sum (ProbData) !=0)

 

Greetings all,

 

I've been playing with BIOMOD for a while and think I have it all sorted out, but I see that this error has come up before. I'm running a species projection with all available model types for a total of 21 species. The first three species run smoothly, but I hit this familiar error:

 

Input:

Projection(Proj=proj, Proj.name='FIAusa', GLM=T, GBM=T, GAM=T, CTA=T, ANN=T, SRE=T, FDA=T, MARS=T, RF=T, BinRoc=T, BinKappa=T, BinTSS=T, FiltRoc=T, FiltKappa=T, FiltTSS=T, repetition.models=T)

 

Output (after listing 3 successful species projections):

Error in if (sum(ProbData) != 0) { :

  missing value where TRUE/FALSE needed

 

If there’s a code setting or otherwise that might be generating this, please do let me know. Just to clarify, I’m not running an ensemble yet, just a regular projection. I would normally be scouring my data file for an issue, but since the first few species run fine, I’m confident that it’s clean.

 

R is running on a 64-bit Linux machine, if that makes any difference.

 

Thanks everyone!

 

-          David

 

________________________

 

David Roberts

PhD Student

Dept. of Renewable Resources

University of Alberta

(780) 492-2540

drr3 at ualberta.ca

 

_______________________________________________






 
Biomod-commits mailing list






 
Biomod-commits at lists.r-forge.r-project.org






 
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/biomod-commits






 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/biomod-commits/attachments/20100621/12a06ecc/attachment-0001.htm>