On 16/04/19 16:50, Jamille Haarloo wrote:
Replacing
ids <- rownames(features)
with
ids <- rownames(predicted)
is the only edit I did after the previous try, so this should have
solved the error.
Watch out: I'm not sure that this is really equivalent. I think I used
the rownames(features) to get the original ids of the features, while
rownames(predicted) apparently gives 1,2,3,4,etc. So you will not be
able to link the results back to your original features...
If I understood correctly na.action = na.exclude can help to work around
the NA values without deleting rows but somehow I did not work. The user
can always compare the original data rows to the results, right? As in
my case, comparing the results file to the vector file shows that 17849
of 17851 segments were classified as expected from the error. I did not
loose any original data and also saved a copy. If you mention in the
manual that intermediate na values might block the analyses and will
therefore be omitted from the final results, it should be perfectly fine.
You are right that na.exclude is probably better here than na.omit.
IIUC, this will give you NAs in the prediction output as well.
So I think that na.exclude is probably the safest bet and shouldn't
create too much of a surprise for the user.
I noticed a minor issue; not all results were added to the vector file -
about 686 segments (almost 4% of the data) were somehow missed.
That's weird. Maybe this is due to the different ids ?
Moritz
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/grass-user