[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
yet another Gstat distribution post!
Edzer's absolutely right; the normal score and the back-transform are not simple
procedures, and there are a number of real problems, such as ties and dealing with
small samples, so that the tails get shortened.
Maybe what we need to decide is whether the complications introduced by
normalizing are justified by the result. When we use non-normal data in
simulation, we are not matching the assumptions of the Multigaussian model, which
is troublesome. I would be nervous about publishing work that used such
simulations. Of course, I am usually nervous about geostatistics in general!
This has been an interesting and (to me at least!) useful discussion, and I hope
others on the list have given some thought to the issue as well. I absolutely
agree that whether or not normalization is implemented in Gstat, its users need to
be aware of what happens (and what does not) in processing.
> I have hardly problems with the normal score transform (although I
> would print a BIG warning in case of ties; also you can decide to
> make ties come out as ties, or assign arbitrary ranks to them (as
> GSLIB does apparently) -- this is also a model decision). I consider
> the back-transform as very non-trivial, one that calls for quite
> a bit of skill, and knowledge how to do this. (As far as I can see
> now, you need to dive into the source code to see what GSLIB really
> does here).
> Implementing may not be the problem -- documenting may be.