[R-sig-Geo] semivariogram + coordinate units

[Pedro Mardones]

Thanks for the advice. I found this comment on a web page (from a
geostats class): "...If the data coordinates were in different units
then we would need to standardize these coordinates (otherwise km in
the horizontal and m in the vertical produce very flatten out
grids)...", so I'm wondering that if by using "relative" locations
(say 0 to 1) instead of the absolute ones (cm or m) can be utilized as
an alternative for the analysis. The point is that by transforming
everything to meters I'm having some problems to fit the models and I
guess is due to the reduced range of the x-axis (0-0.02 m) compare to
the y-axis (0-20 m).
BTW: the grid represents a set of data points measured within a pole,
so that's the reason of the differences in scale (x is from the center
of the pole to the outer part of the pole) and y from the bottom to
the top.
PM



On Fri, Apr 11, 2008 at 9:22 AM, Steve Friedman
<friedman.steve at gmail.com> wrote:
> Pedro,
>
> You absolutely need to have the coordinates in the same units of
> measurement. The geostatistics model is based on the premise of location,
> distance between samples, and the variance between locations.  The bigger
> concern that I can think of is the "change in support"  For example lets
> hypothesize that the variance changes much faster in the x direction which
> in your case is measured in cm.  So far that seems ok, but how can you tell
> whether the variance is changing at the same spatial lag in the y direction
> which is measured (and perhaps sampled) 10 more crudely?  Changing the
> coordinates to the same units of measurement is a transformation that may or
> may not mean much if your sampling scheme is problematic.
>
> Good luck
> Steve
>
>
>
> On Thu, Apr 10, 2008 at 9:19 PM, Pedro Mardones <mardones.p at gmail.com>
> wrote:
> >
> >
> >
> > Dear list members;
> > I'm a newbie in geostatistics so this question may have an obvious
> > answer that I'm not aware of (sorry about this). I've a grid of points
> > in which the x-axis is measured in cm (from 1 to 20) and the y-axis in
> > meters (from 1 to 20). So basically I have 20 x 20 points to work
> > with. Here is where I'm kind of confused. Do I need to transform the
> > coordinates to the same units, say cm or meters, before obtaining an
> > empirical semivariogram? What could be the effect of using the
> > coordinates in the given units (ie without transforming them to the
> > same units) on the analysis?
> > Thanks for any hint
> > PM
> >
> > _______________________________________________
> > R-sig-Geo mailing list
> > R-sig-Geo at stat.math.ethz.ch
> > https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> >
>
>
>
> --
> Steve Friedman
> Phone 561 - 744 - 4642
> Cell 517 - 648 - 6290