Re: [Gnuplot-info] Gnuplot "fit" command

[Tammo H. <Tam...@ie...>]

The attachment didn't make it.
What is the range for x?
If all number differ to much gnuplot will not find a good fit.
You might have to play with scaling factors.

Tammo



On 5 Aug 2006 at 10:34, ra...@be... wrote:

> I was looking for a good, simple program that will let me fit some data to
> theoretical models. I decided to have a look at gnuplot and test it by
> seeing if it can perform a simple linear fit. With the file attached
> (testfile.csv), which fits y=0.4*x, I executed
> 
> gnuplot> f(x) = a*x
> gnuplot> fit f(x) "testfile.csv" via a
> 
> I got the result
> 
> Iteration 3
> WSSR        : 4.13793           delta(WSSR)/WSSR   : -4.55122e-09
> delta(WSSR) : -1.88326e-08      limit for stopping : 1e-05
> lambda    : 0.00822598
> 
> resultant parameter values
> 
> a               = 1.10345
> 
> After 3 iterations the fit converged.
> final sum of squares of residuals : 4.13793
> rel. change during last iteration : -4.55122e-09
> 
> degrees of freedom (ndf) : 14
> rms of residuals      (stdfit) = sqrt(WSSR/ndf)      : 0.54366
> variance of residuals (reduced chisquare) = WSSR/ndf : 0.295567
> 
> Final set of parameters            Asymptotic Standard Error
> =======================            ==========================
> 
> a               = 1.10345          +/- 0.01706      (1.546%)
> 
> 
> correlation matrix of the fit parameters:
> 
>                a
> a               1.000
> 
> 
> 
> Either I am missing something, or there is something very wrong with
> gnuplot's fit command. The data should fit y=0.4*x exactly. I've tried
> setting a to 0.43 before running the fit command and gotten the same
> result.
> 
> Any help?
> 
> - Jonathan