Transforming data help

[Jbryant61]

Can anyone offer advice on how to transform one data set into another.

Thanks
Jason

[RichardJ]

When you say you want to transform one curve to match the other "as best it can", what do you mean by "best"? Without some constraints I can easily make it match exactly, even including the noise.

Richard

[Jbryant61]

Can I see this?
thanks
Jason

[RichardJ]

Divide red by blue. Throw out the points you don't care about. Fit a spline through it. Multiply the blue curve by the spline.

If you replace the spline by a polynomial you can adjust the degree of the polynomial to get it close, but without the noise.

Richard

[Jbryant61]

OK I see. I was really after a modification to the blue curve (scalar as well as a an argument scalar).

Jason

[RichardJ]

On 4/1/2009 9:46:36 AM, Jbryant61 wrote:
>scalar as well as a an
>argument scalar).

I don't understand what you mean. Do you want some wavelength dependent scaling function? Is you do, the polynomial will do that.

Richard

[Jbryant61]

Yes thats correct. So I can extract the polynomial from the spline - is that what you are suggesting?

Jason

[RichardJ]

No. replace the spline with a polynomial.

Divide red by blue. Throw out the points you don't care about. Fit a polynomial through the remaining points. Multiply the blue curve by the polynomial.

Richard

[RichardJ]

I assume the red spectrum is somehow the "real" dye spectrum, complete with lots of shot noise?

Richard

[ptc-1368288]

On 4/1/2009 10:11:01 AM, Jbryant61 wrote:
>Yes thats correct. So I can
>extract the polynomial from
>the spline - is that what you
>are suggesting?
>
>Jason
______________________________

I perceive what you are after, reading Richard. It seems a curve fitting, but can't open my Mathcad today because of the April 1rst virus, I have a fair idea where it comes from, last year the sender read me in the collab. They are happy, they have tested it but don't want to have it today. Simply I hope they will be honest enough not to change the date. Who else than Mathcad cracks could un-register my Mathcad install directory.

jmG

[Jbryant61]

Sorry to hear the news Jean!
Jason

[ptc-1368288]

On 4/1/2009 4:23:14 AM, Jbryant61 wrote:
>Can anyone offer advice on how
>to transform one data set into
>another.
>
>Thanks
>Jason
__________________________

The 1/1 replace part of, is the immediate approach. If you then want to remove noise (filter), the question is to be evaluated by experimenting: before or after replacing ? also it means experimenting lot of filters, like Fourier windows... etc... before/after replacing. The filter may not be available in Mathcad.

Attached is the red right leg smoothed ksmooth bw = 9
Not bad at all as it preserves the principal ondulation, that you can adjust by bw (bandwidth of the ksmooth). That's as far I understand matching. Remember: splining noisy data set is disaster, it means wrongly understanding and using splines.
You can apply the same technique on the left leg, not done. Conclusion: the april 1rst virus, was too small fish to fry ! There was a warning here on the news. Lots of peoples got little birds flying across the screen, not as bad as Peddy the Parrot from Richard !

Salut Jason, Jean.