On Jul 24, 8:36=A0pm, merr...@[EMAIL PROTECTED]
(Ethan Merritt) wrote:
> In article
<7388ede8-9e5a-4280-b704-f7411d695...@[EMAIL PROTECTED]
>,
>
> =A0<goog...@[EMAIL PROTECTED]
> wrote:
> >This seems like yet another user looking for some kind of averaging
> >type smoothing and only finding csplines.
>
> >I believe this is a fundemental flaw in gnuplot.
> [snip]
> >What I guess the user was looking for was some smoothing that would
> >give less weight to the data lying away from the fairly clearly
> >defined straight line, not exagerate the excusions.
>
> >Sadly this does not currently exist.
>
> =A0 You do realize that gnuplot has an entire subsystem, "fit",
> =A0 that will fit any sort of smooth curve you want? =A0Yes, you have
> =A0 to first decide what is the mathematical form of this curve,
> =A0 but isn't that exactly what you are asking for?
>
>
>
> >I think some such smoothing should be provided or (if the reply is
> >that gnuplot is not for data processing) then csplines should be
> >removed.
>
> >regards.
>
> --
> Ethan A Merritt
Thanks Ethan. Yes I do realise about fit command and the way I would
treat the above data would probably be to fit a straight line but that
is hypothetical without knowing what the data represents.
Fit is fine if there is a prototype function to fit , clearly that is
not always the case.
All that does not really address the basic issue I raised, that
fitting a spline through every point is often not what is required.
There are often data points we would prefer to play down rather than
force an excursion to include.
Fitting a spline to experimental data necessarily containing errors
just does not make sense, as plotting the above data shows.
Clearly what the user was expecting was some kind of mean to smooth
the data. This would not be any harder to compute than fitting a
cspline.
I don't understand the inclusion of splines and the exclusion of other
more standard smoothing techniques.
Thanks for your reply.
|
