How to estimate a function that fit well with a curve

[gibnem]

hello
somme one know how to get the function that fit well with a curve
following is my curve:
[PLAIN]http://img814.imageshack.us/img814/5885/tempdepth.png

thx..

 

[pmsrw3]

http://en.wikipedia.org/wiki/LOESS" ?

 

[gibnem]

yeh it could be...

i could create a function that fit very well withe that graph just in excel with a fourth polynomiale and i get the numerical expresion
but withe some other graphe with more variation the polynomiale function don't fit very good so I'm looking if i can found a softwar thar use other methodes (moving average, exp, or a mix of this methodes) to get a better curve fitting)

 

[pmsrw3]

A polynomial is not going to deal well with either the nearly vertical part at high depths, or the nearly horizontal part at low depths. You might be able to fix that with an appropriate coordinate transformation, but a generic method like LOESS will likely produce better results with less effort.

 

[hotvette]

Here is a link to an Excel add-in that claims to have LOESS capability.

http://www.fileguru.com/Data-Curve-Fit-Creator-Add-in/download

If I were doing this I'd consider some form of parametric cubic polynomials (e.g. Bezier) because they can handle vertical slopes.

 

[gibnem]

oooh thank you hotvette
that give a really a very very good fitting