For Programmers: Free Programming Magazines  


Home > Archive > Matlab > December 2006 > parabolic fit









You are viewing an archived Text-only version of the thread. To view this thread in it's original format and/or if you want to reply to this thread please [click here]

 

Author parabolic fit
Ken

2006-12-27, 7:09 pm

Hi,
I have a vector and I would like to fit a parabola to it, and to
evaluateif it is a good fit.
Anyone?
Thank you
John D'Errico

2006-12-27, 7:09 pm

Ken wrote:
>
>
> Hi,
> I have a vector and I would like to fit a parabola to it, and to
> evaluateif it is a good fit.
> Anyone?
> Thank you


polyfit (also polyfitn) will do
the fit equally well.

<http://www.mathworks.com/matlabcent...objectType=FILE>

Evaluate if its a good fit?

Polyfitn returns R^2 and standard
errors for the parameters, as well
as RMS. These parameters may help
to indicate the quality of fit, but
I'd NEVER rely on them. The only
quality of fit that matters is if
the fit is adequate for your
purposes. Only you know what size
the noise is likely to be for your
data. Only you know how good the
fit needs to be. Only you know
other information about the data,
like should the model be monotone
over the domain of the data, but
the curve fit is not.

HTH,
John D'Errico
Ken

2006-12-27, 7:09 pm

Thank you John!
I'll try to use it. I need an automatic process for evaluation
because I need the machine to identify a certain signal based on 2
criteria:
1. tresh hold.
2. Shape.
This is repeated for ~thousands of times so it is must be automatic.
Thank you again.

John D'Errico wrote:
>
>
> Ken wrote:
to[color=darkred]
>
> polyfit (also polyfitn) will do
> the fit equally well.
>
> <http://www.mathworks.com/matlabcent...objectType=FILE>
>
> Evaluate if its a good fit?
>
> Polyfitn returns R^2 and standard
> errors for the parameters, as well
> as RMS. These parameters may help
> to indicate the quality of fit, but
> I'd NEVER rely on them. The only
> quality of fit that matters is if
> the fit is adequate for your
> purposes. Only you know what size
> the noise is likely to be for your
> data. Only you know how good the
> fit needs to be. Only you know
> other information about the data,
> like should the model be monotone
> over the domain of the data, but
> the curve fit is not.
>
> HTH,
> John D'Errico

John D'Errico

2006-12-27, 7:09 pm

In article <ef4986c.1@webcrossing.raydaftYaTP>, Ken <wt_mut@yahoo.com> wrote:

> Thank you John!
> I'll try to use it. I need an automatic process for evaluation
> because I need the machine to identify a certain signal based on 2
> criteria:
> 1. tresh hold.
> 2. Shape.
> This is repeated for ~thousands of times so it is must be automatic.
> Thank you again.


Black box operation is always dangerous.

That said, if you insist on doing so, try
to gain some experience on what is a
reasonable value for parameters like R^2,
rmse, all for your specific problem. These
numbers are meaningless in a vacuum, but
on a consistent group of data sets you
could use them to flag the bad sets.

John



--
The best material model of a cat is another, or preferably the same, cat.
A. Rosenblueth, Philosophy of Science, 1945

Those who can't laugh at themselves leave the job to others.
Anonymous
Sponsored Links







Also available: Server administration forum archive | Web Design forum archive | Software forum archive | Hardware reviews archive

Copyright 2008 codecomments.com