normalize data using continuum task

Guest wrote on Apr 16, 2009

Hello all,

I have a couple of spectra with very low counts in the low wavelength region. I normalize using continuum task in the specred package. In the normalized spectra, the parts with low counts appear exaggerated (oscillate wildly) wrt the rest. Maybe because they were divided by a low value for the fitted continuum. I set my low_rej = 2 and high_rej =5. I fit with a spline3 function of order 1. The higher the fitting order, the worse it becomes. Any ideas to fix this?

Thank you.

Siti :)

---

delta_kuadrant wrote on Apr 16, 2009

hello siti :)

I don't know if this step below can work on your case
but, can you try to exclude the noisy part? you can select the region you want to fit with 's' keystroke before fitting.

CMIIW
regards

Agus Triono P.J

---

Guest wrote on Apr 16, 2009

Hi Agus,

The s key does make it look less noisier but still not as cute as the kitten. Anyway, thanks for your suggestion.

8) Siti

---

delta_kuadrant wrote on Apr 16, 2009

did you try to increase the fitting order?5 or 10 perhaps?
the things you have to consider are rms value of fitting and number of rejection points (CMIIW).
in most cases, if you have small enough rms value and have a couple hundred or thousands number of rejection points, it should be sufficient.
the noisy parts probably also came from previous step, so make sure that your previous steps was well-carried.

CMIIW

Agus

---

Guest wrote on Apr 16, 2009

Dear Agus,

i just want to fit the overall shape of the continuum and not the details of the lines. Increasing the fitting order will fit the sharp drop at the edge because of the flexibility of the spline3 function. I will see how changing the number of rejection points can help.

Thanks for your ideas.

Cheers,

Siti

---

Last post on Apr 16, 2009