[GRASSLIST:1687] Re: m.clump/r.statistics/ and more

[Markus Neteler]

Hi Rado,

On Sat, Mar 31, 2001 at 02:29:54PM -0600, Rado Bonk wrote:
> Hi GRASSusers,
> 
> I found m.clump command in the "Manuals" section of GRASS web page. But I
> did not find it in my compiled GRASS5beta8 on IRIX, neither I did in
> GRASS5beta11 binaries for Linux. Is it a part of GRASS4.3 version?

please give r.clump a try.

> I wonder if there is a way to use GRASS to divide values of slope angle,
> or some other attributes, into statistically acceptable and separable
> ranges or scales. Something like what clustering algorithm do for you in
> ERDAS/Imagine. It is always a problem when one have to categorize
> morphometric property into the certain classes and one does't want to use
> empiricial tresholds.
You may use the GRASS/R interface: Then you have plenty of functions
available. It can store GRASS data (raster/vector/sites) back and forth from
and to R. For those not familiar with "R": It is a free statistics package,
a sort of "S-Plus" clone.
http://www.geog.uni-hannover.de/grass/statsgrass/

> PROBLEMS WITH r.statistics:
> As for r.statistics, is there any limitations on number of region for base
> map? My base map has around 8000 unigue regions, and r.statistics doesn't
> work well. For 5-7% of all regions it produces the desired statistics in
> hexadecimal format (I think it is hexadecimal). When I run:
> r.mapcalc=outputmap_cats=@outputmap
> 
> r.mapcalc dies on "Segmentation fault" on IRIX6.5 machines. Original base
> map was produced by r.clump, and I ran r.support before I used it for
> r.statistics.
> 
> r.statistics doesn't work if method=average. I assume it uses some
> procedures from r.stats, and there is a bug, at least on IRIX6.5. If
> method=min or method=max, it just work for first region in base map(lowest
> categoy value, concerns Linux, and IRIX6.5). Again, the base map has 8000
> unique regions.
Mhhh, please report this as bug in

http://www.geog.uni-hannover.de/grass/bugtracking/bugreport.html

Then we'll look into this.

Kind regards

 Markus Neteler