[R-sig-Geo] Spatial data analysis in MATLAB / Comparison of MATLAB vs R

[Roger Bivand]

On Tue, 20 Jan 2009, Tomislav Hengl wrote:

>
> Dear Barry,
>
> Thanks for your reply.
>
> Funnily enough, I got the same reply from the module coordinator - "you 
> will enjoy learning MATLAB".
>
>
> I have accepted to do everything in MATLAB (I do not have much choice 
> really). Then, I might also try Scilab afterwards and controlling Scilab 
> from R. I will keep you informed about what I discovered about MATLAB 
> (at least considering the spatial analysis capabilities).

Tom,

Can you try going in through the back door using R as a compute server 
from Matlab (or Arc?)? I assume that you are obliged to use Windows, but 
that does get you the R(D)COM opportunity:

http://learnserver.csd.univie.ac.at/rcomwiki/doku.php?id=r_d_com_rcom_and_and_other_software_systems

If Matlab is willing to be a client, then you can hand over some things to 
R from there. Maybe the R.matlab package can also be used with socket 
connections instead of DCOM. In fact, the more able students might find 
Python programming liberating, as I guess you can drive both Matlab and R 
as servers from a Python client - and Python is the currently prefered 
scripting language for ESRI, I believe. The key problem is representing 
objects defined in one language in others, so here points and rasters are 
OK, but lines and polygons are hard.

Note that OSGeo does Python pretty well now too - in terms of how much 
development you'll have to do to prepare the course, you may find that 
there are more packages ready for sticking together with a little language 
than in Matlab itself, and which provide a higher level of abstraction 
(richer objects) than Matlab offers. Matlab is very powerful if objects 
can be readily coerced to matrices, for example in modelling (not fitting 
models but building and calibrating systems models) - there is a different 
view of the world between S/R and Matlab, in GIScience terms a different 
ontology. The priors (model knowledge) probably have more weight among 
Matlab users, the observed data among S/R users (as a rough guess). 
Neither are as good at designing objects or programming - ideally, all 
three would suit your students.

Roger

>
> The real problem is that students want to get training in ESRI and 
> Mathworks products, because they think that it gives them better chances 
> to find work in large/government companies (and this is still largely 
> true). So I think that we really need to 'work' on government agencies.
>
> At least when I look at the Google trend graphs, I can see that some of 
> the commercial players are irreversibly flowing into a decay function :)
>
>
> Tom Hengl
> http://spatial-analyst.net
>
>
>> -----Original Message-----
>> From: b.rowlingson at googlemail.com [mailto:b.rowlingson at googlemail.com] On Behalf Of Barry
>> Rowlingson
>> Sent: Tuesday, January 20, 2009 10:17 AM
>> To: Tomislav Hengl
>> Cc: r-sig-geo at stat.math.ethz.ch
>> Subject: Re: [R-sig-Geo] Spatial data analysis in MATLAB / Comparison of MATLAB vs R
>>
>> 2009/1/20 Tomislav Hengl <T.Hengl at uva.nl>:
>>
>>> I am preparing to teach spatial data analysis at BSc level 
>>> (environmental and Earth sciences). The study programme is completely 
>>> based on MATLAB, which means that I will also need to adjust (I do not 
>>> have much experience with MATLAB).
>>
>>  If you convert your entire study programme to R you'll annoy your
>> colleagues who have to learn and rewrite their courses in R. If you
>> just convert your spatial data analysis module to R you'll annoy your
>> students who'll have to learn R for this and MATLAB for everything
>> else. You choose :)
>>
>>  Ideally all your colleagues will take to R and feel joy at the
>> prospect of learning new software and rewriting course notes - but how
>> often does that happen?
>>
>>  As a first step you could try introducing Scilab or Octave to the
>> students so they can use a MATLAB-ish package at zero cost.
>>
>>  For our undergrad maths course they use Scilab for linear
>> algebra-type stuff, R for stats, and Maxima for computer algebra. You
>> may need to really rework your programme if you want to use just R
>> throughout. It could be worth it.
>>
>>  Don't forget the big argument. R is free and open source - it can,
>> like the research we publish, be freely shared. For me that's
>> ultimately compelling - if it's not open, it's not science; and if
>> it's not science, it's not a BSc.
>>
>> Barry
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no