Re: linear regression

[Robert A. Macy]

Myriam,

for linear well behaved functions, I use...
>> coeff = polyfit(x, y, 1);
with x and y being same size vectors. 

ols.m was mentioned and is also in 2.1.50a binary version
of octave, but I've never tried it.  

If x and y both contain noise, and/or are not monotomic,
with some help from people here I wrote  my own linear
regression called:
 
  splatterfit.m

It returns the same coefficients [well inversed
coefficients] whether x or y are interchanged, In other
words a line through your data set is exact.

        - Robert -

On Mon, 24 Apr 2006 15:43:41 -0400
 Myriam Abramson <address@hidden> wrote:
> 
> How can I run a simple least-square regression in Octave?
> 
> 
> myriam