Open piyanatk opened 2 months ago
@tamerakassie I have added a skeleton notebook to the monte-carlo-sim
branch. Can you see if you can finish it? :)
You will want to pull
and then checkout
the monte-carlo-sim
branch. Then, please work on that branch, committing and pushing new update to the branch as you need. We will then discuss what you find from the simulation.
One of the question that we have from our analysis is that the sky-subtracted incoherently-averaged (over antennas) noise spectrum (INS) of the MeerKLASS autocorrelation visibility does not produce the expected folded Gaussian distribution. This maybe caused by the fact that we do not have enough number of antennas. Specifically, the incoherent averaging over many baselines, or in our cases antennas, in the SSINS algorithm is what leads to the expected noise distribution following the central limit (CLT) theorem. This is derived for cross-correlation in Wilensky et al. (2019). As we only have ~60 antennas to average over, this may or may not be enough for the CLT to kick in.
In attempt to figure this out, we will perform a simple Monte Carlo simulation by drawing from a folded Gaussian distribution (first with no mock RFI for simplicity) 60 times, averaged over the mock noise spectrum, and then look at the z-score distribution.