[Biomod-commits] Re : Re : Biomod question

Thu Jun 10 10:52:36 CEST 2010

Hi again,

Another possibility for plotting your predictions (without setting the independent data) is to run the Projection() function on your original
dataset. This will work with any calibration procedure. As long as the model exists, simply make a prediction on the whole dataset.
It avoids struggling with getting the right lines to only get a partial map.

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 : Bruno Lafourcade <brunolafourcade at aol.com>
A : lailaliana78 at yahoo.com; biomod-commits at lists.r-forge.r-project.org
Envoyé le : Mercredi, 9 Juin 2010 23:13
Sujet : [Biomod-commits] Re :  Biomod question

Hi Liliana,

Q.1
--------
The reason why the level.plot is used using independent data is because we use pseudo-absences.
In that case, the predictions of the models are only produced for the PA data, i.e. a partial sample
of the original dataset. By entering independent data, the predictions are done on the full dataset
given as independent data which : in our case we gave the full original dataset to obtain full predictions
and the correspondant maps.
I hear you are using PA data as well which is why you bump into data length dissimilarities. You can 
however obtain a map of your predictions using PA data by selecting only the coordinates which correspond
to the PA data used. An object called "Biomod.PA.data" is produced when running the Models() function.
It will give the lines that have been sampled from the original data for each run. You are thus able to reenact  
the correspondant data and associated coordinates. But you will only have a partial map of your original data.

Q.2
--------
I will check on that. There are multiple raster formats available. You can always try to convert them before 
exporting them to GIS.

Q.3
--------
The Ensemble.Forecasting.raster() function has lately been added to the package and takes the outputs from 
Projection.raster() to run. As for the ProjectionBestModel(), the adaptation to rasters hasen't been done yet.
Will soon be done.

Q.4
--------
I don't really get your question ? What do you mean by Raster Calculator ?

Hoping to answer your queries.
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 : Claudia liliana Ballesteros Mejia <lailaliana78 at yahoo.com>
A : biomod-commits at lists.r-forge.r-project.org
Envoyé le : Mercredi, 9 Juin 2010 15:20
Sujet : [Biomod-commits] Biomod question

Dear R-forge-List,

I'm using the package biomod to identify areas of potential distribution of abutterfly. I don't have independent data to test, so I'm using a partition ofthe data 80/20. Maybe I should say as well that I chose to run the Models withonly 1 repetitions but 2 runs of Pseudo-absences, and also that I'm running this models under R-2.9.2 linux. But I have also tried to run it under windows R-2.11. (without much success because of the memory limit issue)

I have 4 questions.
1. In the tutorial for this package is described the function level.plot. It isuseful to plot the results of every model but in the tutorial they ran theModels with "independent data", Is there any possibility to use thisfunction without having to run the models with independent data?

load("pred/Pred_Ag_cing")
level.plot(Pred_Ag_cing[,"RF",1,1],coorxy,title='Ag_cingulata RF')

Error in level.plot(Pred_Ag_cing[, "RF", 1, 1], coorxy, title ="Ag_cingulata RF") : 

 data and coordinates should be of the same length 

length(coorxy$Longitude)
[1] 2697182

length(Pred_Ag_cing[,"RF",1,1])
[1] 1350

It gives me an error saying that the coordinates and the data must be the samelength, (the truth is I don't get why without running the models withindependent data it doesn't get the same length but with independent it does.) 

2. Is there any possibility to export the data to make
the plots as an ascii files so can be imported into GIS? I know there is the function Projection.raster (under windows R.2.11) but when I try to imported into GIS it doesn't show anything (does it also exist in Linux?) 

3. When using Projection.raster function, it doesn't produce the output that the functions Ensemble.forecasting and ProjectionBestModel requires to work. But if I use the function Projection is not possible to export them as ascii files. Is it true?, Is there anyway to solve this? 

4. If the only way to plot the outputs of the models in GIS is by extracting themodel and using Raster Calculator, how do I get the final models of RandomForest, MARS and GAM?

Any help would be very much appreciated, 

Thanks a lot in advanced,

Liliana Ballesteros
PhD Student
University of Basel

_______________________________________________

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

_______________________________________________

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/20100610/1d832f0a/attachment.htm>