Message-ID: <AD5CA6183570B54F92AA45CE2619F9B90129DB9E@gi-zrmk.si>
Date: 2011-07-25T20:32:06Z
From: Matevz Pavlic
Subject: Indicator kriging map creation
In-Reply-To: <4E2C8E86.9040308@uni-muenster.de>
Hi Edzer,
sorry for the long time to reply.
>It might be annoying, and after all, it is your sole responsibility to make sure the right labels end up in the right place. R can only prevent you from making mistakes that are too obvious.
What do you think would be the best thing to do about that? Predict the classes on 3D grid? I mean, i would like to create a code that would generate maps of several classes (variing numbers of them) each depth interval. I think that prediction on 3D grid would be the right thing?
>Not so likely; I think it comes with simple kriging -- rare classes often show weak autocorrelation, resulting in low estimates everywhere.
>Without having seen your data, I suspect your classes 9 and 10 are like that.
Attached are the co-variograms of ten classes. As i can see there are few classes (devet=level 10 , osem=level 9...) that have nugget (pure nugget effect). This is probably because there are only a few point with that class in the area :
devet dva ena nic osem pet sedem set stiri tri
5 50 85 209 3 7 83 71 49 89
Is there a way to fix this without discarding this classes?
Regards,
m
-------------- next part --------------
A non-text attachment was scrubbed...
Name: variogram.png
Type: image/png
Size: 26318 bytes
Desc: variogram.png
URL: <https://stat.ethz.ch/pipermail/r-sig-geo/attachments/20110725/487b5c82/attachment.png>