statistics for GIS?

"""It would not be hard to imagine using much of the functionality of
S-Plus, from classification and regression trees to data visualization.
This would be a significant addition to the usefulness of GRASS."""

There is no reason to do not use Splus for the statistical analysis
of data coming from Grass. I do it a lot, I would say most of the time.
It requires writing few and simple Splus functions.

i missed the original thread, but the above comments caught my eye.
i currently use S-Plus and GRASS daily. i dump data from grass
into S-Plus, a trivial matter as Agus said. however, the return
trip is not trivial, or if it is, my abilities are less than trivial.
i use the binary-tree capabilities of S-Plus and need to reproduce
the tree structure in a language that GRASS understands, not the
S-Plus object stored and maintained in S-Plus for further tree
analyses and prediction. S-Plus will not handle data sets the size
of which i am using, so i go through the painful and tedious process
of reconstructing the final trees in r.mapcalc. an interface between
S-Plus and GRASS would presumably eliminate this step. i currently
plan to switch to arc/info since i understand that it does talk to S-Plus,
or will in the near future. i would rather stick with grass.

venables and ripley's new book on S-Plus is outstanding and includes
even more spatial statistics that are a gis dream. they
supplement mccauley's previous contributions, and are likely to
get more use and support given mccauley's overloaded work schedule.
at this point i feel S-Plus is SO valuable that it makes more sense
to switch gis's than mathematical packages.

if you read between the lines here, my programing abilities a poor,
and someone else could surely write a routine to transfer the tree
objects into GRASS. has anybody out there done this? the bigger
picture and the more pertinent question (that i probably missed in
the original message) is whether an interface is planned that would
allow use of all of the current S-Plus pearls in GRASS?

kelly

kelly@icess.ucsb.edu

"""however, the return
trip is not trivial, or if it is, my abilities are less than trivial.
i use the binary-tree capabilities of S-Plus and need to reproduce
the tree structure in a language that GRASS understands, not the
S-Plus object stored and maintained in S-Plus for further tree
analyses and prediction. S-Plus will not handle data sets the size
of which i am using, so i go through the painful and tedious process
of reconstructing the final trees in r.mapcalc. an interface between
S-Plus and GRASS would presumably eliminate this step."""

I'm a bit confused about your needs. Could you provide an example?
What I normaly do is to use Grass to generate data files according to
GIS (grass) rules. Then I Use Splus over those files and bring the
results to grass again for display and further GIS processing. To
import/export to/from Splus I use simple functions that are based
on write and scan Splus commands. The most "sophisticated" (but still
very simple) is the function that I wrote to import binary (cell)
grass files to Splus (and the equivalent to export it).

"""i currently
plan to switch to arc/info since i understand that it does talk to S-Plus,
or will in the near future."""

Could you detail what kind of "talk" they perform?

Agus