Hi all,
I have a problem with the function raster::predict very similar to the one
described here [1], [2] using raster package version 2.0-41 and party package
version 1.0-6, where my model is a conditional inference forest
(party::cforest). Could not find a solution in either post.
The problem is the following: when I use the function predict() I get an error
relating to a mismatch between the levels of factors in the data frame used to
fit the model and those of the raster layers used to predict to the whole
surface.
The only difference that I can see is that the levels of the factors in the
model are as follows (e.g.):
> morphrec.levels<-levels(v$morphrec)
> morphrec.levels
[1] "1" "2" "3" "4" "5" "6"
Whereas if I ask for the unique values of the same raster layer (where values
were extracted from in an earlier step) that is now being used to predict a
value for the response variable, I see the following:
> morphrec.values<-sort(unique(getValues(morphrec)))
> morphrec.values
[1] 1 2 3 4 5 6
When I try running predict, the following happens:
r <- predict(predictors, confor.dens, type='response',OOB=TRUE)
Loading required package: tcltk
Loading Tcl/Tk interface ... done Error in checkData(oldData, RET) :
Classes of new data do not match original data
> predictors
class : RasterStack
dimensions : 2787, 2293, 6390591, 7 (nrow, ncol, ncell, nlayers) resolution
: 200, 200 (x, y)
extent : 296387.5, 754987.5, 7488413, 8045813 (xmin, xmax, ymin, ymax)
coord. ref. : +proj=utm +zone=33 +datum=WGS84 +ellps=WGS84 +towgs84=0,0,0
names : bathy, slope, tri, grainsize, landscape,
sedenv, morphrec
min values : 0.44850001, 0.02432192, 0.04429007, 1.00000000, 21.00000000,
1.00000000, 1.00000000
max values : 2912.94849, 46.13418, 43.59974, 300.00000, 431.00000,
5.00000, 6.00000
> confor.dens
Random Forest using Conditional Inference Trees
Number of trees: 1000
Response: density
Inputs: bathy, slope, tri, grainsize, landscape, sedenv, morphrec Number of
observations: 1020
Is there any way of telling predict() that the values of some raster layers
will have to be considered characters? Or am I facing a different problem
entirely?
Any help will be much appreciated. I have been working on creating a small,
reproducible example, but I can't seem to get the exact same behavior.
Regards,
Geno
[1]
http://r-sig-geo.2731867.n2.nabble.com/problems-with-predict-in-package-raster-td6627291.html#a6628041
[2]
http://r-sig-geo.2731867.n2.nabble.com/new-function-predict-package-raster-td5796744.html#a5816769
Genoveva Gonzalez Mirelis, Scientist
Institute of Marine Research
Nordnesgaten 50
5005 Bergen, Norway
+47 5523 6376
[[alternative HTML version deleted]]
_______________________________________________
R-sig-Geo mailing list
R-sig-Geo@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
