Tutorial Questions - Day 2

Open Data Workshop
1

Questions for Day 2 tutorials (May 16, 2023) can be posted here

2

How can we construct our own waveform templates and use them here ?

1 Like
3

Hi @GBRK5, are you asking how to use the pycbc waveforms with different parameters? If you mean how you can build your own waveform model and upload it to pycbc, this is a rather complicated process, depending on the accuracy you want to achieve. There are different approaches depending on the system you want to study, Effective one body waveforms https://arxiv.org/pdf/gr-qc/0103018.pdf, Numerical relativity informed waveforms https://arxiv.org/pdf/1508.07250.pdf and Analytical waveforms based on post-newtonian taylor expansions https://arxiv.org/pdf/1606.03952.pdf etc.

4

You can also create your own time domain and frequency domain model and use for parameter estimation although it is a bit involved. Here is a simple exercise using bilby examples/gw_examples/injection_examples/create_your_own_time_domain_source_model.py · master · lscsoft / bilby · GitLab

5

Hi, I do not get the need of normalizing the times over the sample rate as
times = numpy.arange(len(data) - len(hp1_numpy)) / float(sample_rate) in the cell before the “Detection in Colored Noise” In tutorial 2.1, can you please explain? Thanks!

6

This is to rescale the times axis in the plot

7

Yes, thanks, but why it is needed?

8

This is actually to get the get the merger at the exact point. Earlier you generated the event at a random point.

9

Tutorial 2.1

When we inject the one component of the GW(plus polarization) signal into the white noise data, can I calculate the SNR of the injected signal?

10

Yes. Using the matched filtering technique details of which are given in tutorial 2.2

11

Thanks for pointing to the paper.

Is there a model independent way of characterising the noise? Model agnostic and non-gaussian

12

In tutorial 2.3,

The units of the PSD plot is Strain^2/Hz.

Should it not be 1/Hz^2 Power spectrum?

13

PSD (Power Spectral Density) is already a power, so 1/Hz is correct; ASD (Amplitude Spectral Density) has unit of measurement equal to 1/sqrt(Hz) instead

1 Like
14

(As far as I know), you have anyway to assume a model of your noise for doing signal inference. Globally, the Gaussian assumption is “almost” correct (also in plotting the PSD, doing the Fourier Transform you are assuming that your data are overall stationary), or at least it allows you anyway to analyze your signal

1 Like
15

Hi All, are there links to the presentation files (PDF) etc? They were mentioned but I could not find them.

1 Like
16

Hi, I want to know whether the Colab works after the workshop finished?

17

Do we need to submit all the assignments of all three tutorials by today (17 May) itself? Or can we submit it till tomorrow (18 May) morning?..

1 Like
18

Yes @leesong , you can always access the tutorials and colab too as it is an open source resource.

19

Hi @Anushka you can access the thinkific course and complete it whenever you want, there is no deadline. To participate in the data challenge you have one week time if you want it to be evaluated.

20

I don’t understand what we are doing in Tutorial 2.2 in the section " Calculate the Power Spectral Density". I understood that we want to use 1/PSD as a filter, but isn’t that the same as data whitening? What does “interpolate” do?