Thread Subject:

find y=f(x)

Hi,

I have a curve (I don't know it's equation) and I want to find Y such that Y=f(x) where x is given. Is there a way I can do it without having to fit the curve with polynomial approximation? Because when I fit the curve with polynomial approximation, it gives me curves but with a completely different shape...

Thanks a lot.

"S?bastien Vilaseca" <seb2000@bluewin.ch> wrote in message <gp3jku$hf1$1@fred.mathworks.com>...
> Hi,
>
> I have a curve (I don't know it's equation) and I want to find Y such that Y=f(x) where x is given. Is there a way I can do it without having to fit the curve with polynomial approximation? Because when I fit the curve with polynomial approximation, it gives me curves but with a completely different shape...
>
> Thanks a lot.


The problem is unclear: do you have a set of data and you want to fit a curve?
You know your data are generated by a certain function that you don't know and you to infer?

Please be more specific.

"Lorenzo Guerrasio" <lorenzo.guerrasio@email.it> wrote in message <gp3no5$p9m$1@fred.mathworks.com>...
> "S?bastien Vilaseca" <seb2000@bluewin.ch> wrote in message <gp3jku$hf1$1@fred.mathworks.com>...
> > Hi,
> >
> > I have a curve (I don't know it's equation) and I want to find Y such that Y=f(x) where x is given. Is there a way I can do it without having to fit the curve with polynomial approximation? Because when I fit the curve with polynomial approximation, it gives me curves but with a completely different shape...
> >
> > Thanks a lot.
>
>
> The problem is unclear: do you have a set of data and you want to fit a curve?
> You know your data are generated by a certain function that you don't know and you to infer?
>
> Please be more specific.

Yes, I have a set of 10 datapoints (x,y). I plotted the curve that rely those points. What I want to know is the value of y for EVERY x, not only for those that I have in my dataset.

Thanks.

"S?bastien Vilaseca" <seb2000@bluewin.ch> wrote in message <gp3q19$48$1@fred.mathworks.com>...
> "Lorenzo Guerrasio" <lorenzo.guerrasio@email.it> wrote in message <gp3no5$p9m$1@fred.mathworks.com>...
> > "S?bastien Vilaseca" <seb2000@bluewin.ch> wrote in message <gp3jku$hf1$1@fred.mathworks.com>...
> > > Hi,
> > >
> > > I have a curve (I don't know it's equation) and I want to find Y such that Y=f(x) where x is given. Is there a way I can do it without having to fit the curve with polynomial approximation? Because when I fit the curve with polynomial approximation, it gives me curves but with a completely different shape...
> > >
> > > Thanks a lot.
> >
> >
> > The problem is unclear: do you have a set of data and you want to fit a curve?
> > You know your data are generated by a certain function that you don't know and you to infer?
> >
> > Please be more specific.
>
> Yes, I have a set of 10 datapoints (x,y). I plotted the curve that rely those points. What I want to know is the value of y for EVERY x, not only for those that I have in my dataset.
>
> Thanks.

Then it is not clear this "it gives me curves but with a completely different shape..."
Different shape with respect to what? If you want your curve to pass throu all the points, you'll maybe interested in help-->interpolation
Otherwise polyfit + polyval, but you should you a proper number of paramenters to fit (not too much,not too few...you have only 10 points, it depends on the noise).

In the formar case, you'll not have an y=f(x), but just a Y(X)...

Regards

"S?bastien Vilaseca" <seb2000@bluewin.ch> wrote in message <gp3jku$hf1$1@fred.mathworks.com>...
> Hi,
>
> I have a curve (I don't know it's equation) and I want to find Y such that Y=f(x) where x is given. Is there a way I can do it without having to fit the curve with polynomial approximation? Because when I fit the curve with polynomial approximation, it gives me curves but with a completely different shape...
>
> Thanks a lot.

The real problem here is that you seem to be expecting a bit much from MATLAB.

If you have a collection of points but no idea of the function that links them then there are any number of methods of fitting a line to those points. Any of these methods will provide a fit of greater or lesser accuracy depending on the amount of freedom you allow the parameter estimation process and the freedom implicit in the equation being fitted.

For example a 10th order polynomial will fit your data perfectly but, as I think you have already experienced, it probably isn't what you're looking for if you intent to interpolate. A straight line will be less uneven but has little freedom so may not fit very well.

The key is to identify a likely function for f(x) then use MATLAB it to parameterise it.

Further reading might include smoothing functions which are pretty much what you seem to be talking about.