2
$\begingroup$

I'm working with data from radio surveys. My images are .fits files with BUNIT Jy/beam. I want to calculate the net flux from a continuous source (not point source). When I simply added all the intensity values and converted it into Jy using the beam parameters, my professor said that it is not enough. He said, "You have to account for noise by taking 3 or 4-sigma level depending on the distribution of the noise". The distribution of noise (at places other than the source) is nearly Gaussian as I can see from a fits viewer.

By "take 3-sigma level", does he mean that I should fit a Gaussian to the noise, find sigma and then choose the pixels with intensity values greater than 3*sigma? If I do this, obviously I lose a lot of pixels. Also, if this is correct, should I subtract the mean of intensity of "noise" pixels from my source intensity?

An RMS noise is given in the header, with a caution that it varies highly around some sources. Is this RMS noise value in anyway related to the sigma that I calculated earlier?

$\endgroup$
  • $\begingroup$ Your professor is the only person who knows exactly what they meant, and we'd just be interpreting that and might be wrong. It is better to ask your professor to clarify the procedure they want. $\endgroup$ – StephenG Sep 14 '17 at 4:33
1
$\begingroup$

Maybe too late but probably he is referring to only take into account pixels with an intensity greater than 3*sigma. If not, you will be counting random noise as continuum. Obviously you will lose a lot of pixels, but those pixels doesn't carry any information at all.

$\endgroup$
1
$\begingroup$

The professor might also mean to subtract the background region fluxes for the noise subtraction. Then, calculating the statistics for the subtracted flux value and error of the measurement, say A +/- dA. Then, applying the 3-sigma criteria to determine whether there is a radio source detected at the location, i.e., if A - 3*dA > 0, there is a source, otherwise there is no source.

If there is no source, you will want to quote the value for 3-sigma upper limit instead, e.g., A + 3*dA.

Also, note that for a faint source, regular poisson or Gaussian statistics are not good representative. In X-ray, we use Cash statistics for low-count cases. I think the same statistics might be applicable with radio observations.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.