Re: [Gnuplot-info] range of fitted variable

[Tatsuro M. <tma...@ya...>]

> Even with Marquardt-Levenberg method, you can constrain parameters.

OK. This is just my ignorance.
Anyway, gnuplot fit uses unconstrained least square method.

> Anyway, you can always does it in the function, and put the constrains here
> since conditional statement are accepted. 

help fit says
gnuplot> help fit
 The `fit` command fits a user-supplied real-valued expression to a set of
 data points, using the nonlinear least-squares Marquardt-Levenberg
 algorithm. There can be up to 12 independent variables, there is always 1
 dependent variable, and any number of parameters can be fitted.
 Optionally, error estimates can be input for weighting the data points.

 The basic use of `fit` is best explained by a simple example:

       f(x) = a + b*x + c*x**2
       fit f(x) 'measured.dat' using 1:2 via a,b,c
       plot 'measured.dat' u 1:2, f(x)

 Syntax:
       fit {<ranges>} <expression>
           '<datafile>' {datafile-modifiers}
           {{unitweights} | {y|xy|z}error | errors <var1>{,<var2>,...}}
           via '<parameter file>' | <var1>{,<var2>,...}


----- Original Message -----

> From: Patrick Dupre 
> To: Tatsuro MATSUOKA 
> Cc: gnuplot 
> Date: 2019/6/30, Sun 07:18
> Subject: Re: [Gnuplot-info] range of fitted variable
> 
> Hello,
> 
> I am not sure that it is a helfull response.
> Even with Marquardt-Levenberg method, you can constrain parameters.
> Anyway, you can always does it in the function, and put the constrains here
> since conditional statement are accepted.
> 
> Regards.
> 
>>  Subject: Re: [Gnuplot-info] range of fitted variable
>> 
>>  --- pdupre
>>  > Hello,
>>  >
>>  > Can I set a range to the fitted variable ?
>>  > An option could be to set a if (condition) in the fitting function. Is 
> it possible?
>>  >
>>  > Thanks
>>  >
>> 
>>  fit of gnuplot use Marquardt-Levenberg method.
>>  Marquardt-Levenberg method is a uncostrained method. You cannot set bounds 
> of parameters.
>> 
>>  Tatsuro
>> 
>> 
>