Re: [Gnuplot-info] points too far from the average not enough considered in the fit procedure

[Hans-Bernhard B. <HBB...@t-...>]

parigigi wrote:

> Particularly if each point weights equal, the fitting procedure "ignores" a
> lot points far from the others, 

No it doesn't.  Being "far from others" bears no relevance at all to how 
a fit is supposed to handle data points.  It's _only_ about how far they 
are from the model function, and then only in the z direction.

> but since I deal with resonances, I am
> interested in the converse: something that weights much more points far from
> the average.

So supply your own weights, instead of letting gnuplot default to them 
all being equal.  The documentation is clear enough about how to do 
that, I think.  If you think otherwise, tell where you got lost.

> Is there a way to change automatically the weight of each point in a way
> that fitted surface will be closer to them?

Of course not!  Doing that automatically would render whatever parameter 
  you might end up extracting completely meaningless.