Continuum fit - Source code

[babbloo]

Hi there

I'm stuck with the continuum fit for the spectra. Even though IRAF has a built in continuum routine, it doesn't work well with all the spectra. So, I'm trying to look into the continuum algorithm to make some minor changes and make it work for all my spectra. But, I cannot find the source code for the built in continuum algorithm. I tried searching it in IRAF website but in vain.

It would be really great if I can get the source code of it from any of you, so that I can analyse it and make some changes to make it work for my data.

Can someone send me the source code?

Thanks

[jturner]

It looks like continuum is just the same task as sfit, with a different name and different parameter defaults. The code is "onedspec$t_sfit.x".

Cheers,

James.

[babbloo]

James

Thank you!
Another quick question ...
As I have mentioned earlier, I'm trying to fit continuum to QSO sepctra ...
But, it doesn look really good for all the spectra ... and changing the parameters for each spectra is no good .... any suggestions about this?

Also, for one spectra the fit looks good if I increase the order to 12 .. with a cubic spline algorithm ... Is it alright if the order is that high ... it was mentioned that for a cubic spline algorithm order denotes the the number of spline pieces .. I really didn understand this clearly ... can you explain?

Thanks
Srinivas

[valdes]

Hi,

Continuum fitting totally automatically is very hard. In other words, there may be not single set of parameters that will work to your satisfaction in all cases. The best that might be done is to have a list of features in wavelength to be excluded. Of course you would need to have all the QSO spectra calibrated to the same wavelength frame.

As for cubic splines... To fit arbitrary functional shapes requires higher orders. However, for polynomial functions this leads to instabilities and possible abrupt behavior, particularly if interpolating/extrapolating across regions of mssing data. A spline is a technique to have a low order polynomial form locally but allow following fairly arbitrary shapes. A Nth spline is always locally a polynomial of Nth order; i.e. for cubic splines it is locally cubic. The idea of "pieces", also related to the term "knots", defines the size of the local regions. Think of fitting a cubic function to each piece but in a way that is continuous across the pieces. Anyway, cubic splines are the recommended method when going beyond shapes of 4th or 5th order unless the data points are fully sampled (no exclusions or deletions or extrapolations).

Yours,
Frank Valdes