On Jun 21, 12:14=A0pm, Ben <benoit.bar...@[EMAIL PROTECTED]
> wrote:
> > For example, the maths of cspline is based on the fact that the line
> > will pass exactly through each point and have a smothly changing
> > derivative (slope).
>
> > If either of these conditions is not true of your data the fitted
> > spline will be a misrepresentation. This is, in fact, very likely for
> > experimental data containing an error or uncertaintly as in usually
> > the case.
>
> My plots have a low value of derivative, but really different shapes.
> The goal is just to make a picture to show a gain or a losse. There
> will be no physcal interpretation with this.
>
> So keep an easy way with table (how haven't I think about this?) seems
> really good.
> Thank you for your answer, table seems really the good solution.
OK, just be aware that splines can over-shoot or undershoot depending
on the data so your difference may be a loss instead of a gain. I
don't know what value you could put on your "picture" made this way.
Since you don't say what your plot represents I can not suggest a
better way than the general suggestion to use fit.
The table method would also apply to fit so maybe you could adapt it
later.
Watch out for splines making things look oscillatory when they are
not.
good luck.
|
