[Date Index] [Thread Index] [Author Index]

Re:

• To: mathgroup at smc.vnet.net
• Subject: [mg6976] Re:
• From: seanross at worldnet.att.net
• Date: Wed, 30 Apr 1997 22:26:15 -0400 (EDT)
• Sender: owner-wri-mathgroup at wolfram.com

Markus Steffan wrote:
> 
> Hello,
> 
> I found your email-address in the newsgroup comp.soft-sys.math.mathematica
> and I would like to know if/how I can do multi-dimensional non-linear
> Gauss regression (least squares) with mathematica (or if this is not possible
> with another program). I hope you can help me with this.
> 
> Regards
> 
>   Markus Steffan.
> 
> ----------------
> 
> email: msteffan at mi.uni-koeln.de

#1:  Mathematica is a high level programming language.  It can do 
anything that can be done in Fortran or C as far as number crunching 
goes.  So the answer to any question like "Can I do .... with 
mathematica?" is yes.  As far as it being possible with mathematica or 
another program: be courageous!  You don't need to look all over the 
world for some canned routine to solve a problem for you.  Learn a 
programming language (or several) and write your own!

#2:  It turns out that a pre-written routine is available in 
Mathematica for what you want.  The command Fit uses least squares 
fitting and will take multidimensional basis functions.  I have done 
fits with Sqrt[x^2 +y^2]and Exp[-x^2-y^2] to various powers.