[Biomod-commits] Noisy binary prediction

[alexandre sampaio]

Hi,

I've produced an ensemble forecast prediction that have a noisy
pattern of dispersion of binary presence points. There are some
continuous areas of predicted presence but most of the prediction
comes on scattered present points dispersed on systematic patterns
like waves (see attached pic). Is this normal? Is this related to the
method to create pseudo-absences? If not what may be causing it?

Thanks for the help!

Alexandre Sampaio

Code:
 Initial.State(Response = back[,7], Explanatory = back[,1:3],
 IndependentResponse = NULL, IndependentExplanatory = NULL,
 sp.name="melinis")

 Models(GLM = T, TypeGLM = "poly", Test = "AIC", GBM = T, No.trees =
 2000, GAM = T,
 Spline = 3, CTA = T, CV.tree = 50, ANN = T, CV.ann = 2, SRE = T,
 Perc025=T, Perc05=F, MDA = T,
 MARS = T, RF = T, NbRunEval = 3, DataSplit = 80, Yweights=NULL, Roc =
 T, Optimized.Threshold.Roc = T,
 Kappa = T, TSS=T, KeepPredIndependent = F, VarImport=F, NbRepPA=1,
 strategy="circles",
 coor=CoorXY, distance=2, nb.absences=1000)

 Projection(Proj = back[,1:3], Proj.name='melinis', GLM = T, GBM = T, GAM =
T,
 CTA = F, ANN = T, SRE = T, MDA =T, MARS = T, RF = T,
 BinRoc = T, BinKappa = T, BinTSS = T, FiltRoc = F, FiltKappa = F, FiltTSS =
F,
 repetition.models=T)

 Ensemble.Forecasting(Proj.name= "melinis", weight.method='Roc',
PCA.median=T,
 binary=T, bin.method='Roc', Test=F, decay=1.6, repetition.models=T)

 level.plot(consensus_melinis_melinis_full_bin[,2], back[,1:3])
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.r-forge.r-project.org/pipermail/biomod-commits/attachments/20100419/e1242ad0/attachment-0001.htm 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sample-tanz.jpg
Type: image/jpeg
Size: 37104 bytes
Desc: not available
Url : http://lists.r-forge.r-project.org/pipermail/biomod-commits/attachments/20100419/e1242ad0/attachment-0001.jpg