Questions on Data Challenge

Hi! Since the deadline is in 2 hours and I’ve completed the challenge, I guess it’s legal to ask questions.

So, for me the first 3 parts were rather simple, but the 4th was challenging and I have a lot of questions on how we could recognise the glitches in the data.

My thoughts were, as I see the high SNR rate on some time with one waveform template on one of the detectors but do not see anything like it on the another – I guess the reason was local and could be considered as a glitch. Am I right? But when I try to plot q-transform and look on the signal if it corresponds to any form of glitches that were presented in the lectures, suddenly, I see no signal! (I am talking about 7-8 seconds signal on L1 data). What could be the reason?

Another questions I had are about the optimal way to search the signals. In the challenge I’ve just tried to search for any high enough SNR-rates over the array from 10 to 50 s.masses with a step of 1 solar mass, but it cost a lot of computational time. So, I know that in continuous gw searches the “mismatch” characteristic and different grids and methods for transient searches could be used, are there any similar ways that could be used for that problem? How generally should I solve the presented problem with glitches and different signals the most optimal way? And how should I recognise if the signal that I’ve found with high enough SNR rate is from an astrophysical source or just a glitch of some sort?

Thank you in advance! The workshop was great and useful and I had a lot of fun solving the tutorials!

@M206265 Thank you for the questions!

This challenge is somewhat open-ended. There’s no 1 “right” way to distinguish signals and glitches.

In general, yes, checking for coincidence between the two detectors is an important step, and is one of the most important cross-checks we have in our search for signals.

Another approach that would work well for this challenge is to plot Q-transforms of each potential signal. For BBH’s, we can typically see the signal as a “chirp” in the Q-transform. Glitches almost never mimic this chirping pattern.

Yet another approach is to calculate a “chi-squared” value for the signals, as described in Tutorial 2.3.

1 Like