[SciPy-user] linear regression

[zunzun at zunzun.com]

On Sat, May 12, 2007 at 12:14:38PM +0200, Stef Mientki wrote:
> 
> But I see that the relation is not completely linear,

Try the 2D function finder at http/zunzun.com, or the
site fitting code at http://sf.net/projects/pythonequations.
Bot scipy and weave are needed to use the source code.


> so I want to split the x-y pairs in 2 sets, one upto the value of 150 
> and and with the values above,
> and would expect that this split up will give better correlations.

Having calibrated a lot of sensors myself, this would seem reasonable.
My work was mostly X-ray and gamma ray stuff - flouresence,
backscatter and absorbtion of monoenergetic and full spectrum.
I also worked with beta ray and optical sensors.


> I'm doing something wrong ?

Send me an example dataset (after you try the site) and I'll
be glad to take a look at it.

     James