Re: variances

[Konstantin Malakhanov <kosta@iwwnt.iwwl.rwth-aachen.de>]

>>>>> "Marlous van der Meer" == Marlous van der Meer <m.vandermeer@geog.uu.nl> writes:

    Marlous van der Meer>  When I looked at
    Marlous van der Meer> one of the krigged maps, I found out that
    Marlous van der Meer> the variance map showed almost only negative
    Marlous van der Meer> values. I don't believe that it is possible
    Marlous van der Meer> for the variance to have a negative value!
    Marlous van der Meer> Can anyone explain this?

    Marlous van der Meer> The script looks as followed: points (HM):
    Marlous van der Meer> 'hoog65a.txt', x=1, y=2, v=3, X=4,
    Marlous van der Meer> radius=3000, min=2, average; variogram (HM):
    Marlous van der Meer> -100 Nug(0) + 150885 Gau(2469,63.0,0.432);
    Marlous van der Meer> mask: 'afstand.map'; predictions (HM):
    Marlous van der Meer> 'tmp.prd'; variances (HM): 'tmp.var';

Well, if YOU put negative variance into play (with -100 Nug(0)), don't 
blame kriging for giving you negative variances back ;-). More
seriously, nugget effect cannot be negative, as it is the variance of
data at equal positions.

Hope it helps.
-- 

Konstantin Malakhanov
---
Institut for Hydraulic Engineering & Groundwater Mgmt., 
RWTH Aachen, Germany
E-Mail: malakhanov@iww.rwth-aachen.de          Tel. +49/(0)241/807343