**Attendance**

Abhirup, Anuradha G., Krishnendu, Nathan, Saleem

**Agenda**

**Action Items:**

- Krishnendu: Consider the performance of the test with injections with the same total SNR but different inspiral lengths (and thus different inspiral SNRs).

- Krishnendu: Update the Git hash on the wiki page to the latest version to be reviewed.

- Krishnendu: Collect pointers to all runs (even ones on older code) that have been performed with this test so far on the wiki and point the reviewers to them (putting the runs with older code on a separate wiki page). This includes the non-BBH runs.

- Krishnendu: Prepare configs on the TGR relations injections and run them past Abhirup and Nathan. 

**Minutes**

Krishnendu’s update: Everyone agreed about the choice of sampling parameters, but the related code changes need to be discussed. Discussing with Saleem how to decide whether to run the test on a given event. Found a reasonable cutoff for applying this test of an inspiral-only SNR of > 10 for O1 and O2 BBH events (defining inspiral as the frequencies below the IMRPhenomD end-of-inspiral dimensionless frequency of 0.018). For the methods paper injections use a cutoff of 12 on this inspiral SNR (see https://git.ligo.org/nv.krishnendu/lalsuite/wikis/Details-of-the-injection-studies) and propose this as the criterion for running this test on real events.

Discussion about whether this is a good choice. Abhirup suggests using the same SNR threshold as TIGER/FTA use. It’s currently 6 for the O2 testing GR paper (which is fairly ad hoc, and didn’t seem to be the best choice a posteriori, as Nathan pointed out), but in the future this threshold could change, and it might make sense for all three tests to use the same threshold. Abhirup mentions that besides SNR one could consider cycles in band. Nathan mentions that Michalis started looking at effective cycles as a potential criterion for deciding to run a test for the O2 testing GR paper, but this didn’t reach a conclusion. Anuradha says that it’s understandable that Krishnendu doesn’t want to do lots more runs on injections, but suggests a small test (see the action item). Abhirup mentions that deciding on this criterion is important for the review.

Krishnendu confirms that the table of code changes on the wiki page is up to date.

Abhirup asks about the status of the run to check that the kappa_1, kappa_2 and kappa_s, kappa_a parameterizations are equivalent. Krishnendu says that she will send details about this run by e-mail.

Anuradha points out that the Git hash to be reviewed given on the wiki page is not the latest one.

Abhirup points out that all the runs in the table linked above have total mass 30 Msun and asks about what other runs have been performed. Krishnendu says that 30 Msun is the maximum total mass they considered for most runs, though they tried runs with a mass up to 80 Msun, though there the cutoff frequency is very low, ~24 Hz. They have previously considered a variety of mass ratios and spins.

Abhirup states that part of the review should be to push the test into regions where it will no longer make sense, showing where it fails. The two regions are high masses and low SNRs for low-mass systems (so both give low inspiral SNRs).

Abhirup asks whether they’ve analyzed any LIGO-Virgo events besides GW151226 and GW170608. Krishnendu says that they have run on all the BBH events. Abhirup asks to have access to all these other runs, so the review can make statements about not applying this test to high-mass events. Krishnendu mentions that the older runs are only with kappa_s runs (with kappa_a = 0), not with kappa_1 and kappa_2, and had a slightly different code setup, but the physics is the same.

Abhirup asks about non-BBH injections: Do they have the infrastructure to make injections with non-BBH kappa parameters? Krishnendu says that they have performed such runs and some results [for (10, 5) Msun binaries] are reported in the methods paper, but have higher-mass runs [(20, 5) Msun], as well. kappa_s was injected from -50 to 50, since the waveform didn’t look good for larger magnitudes of kappa_s.

Saleem mentions that they have made studies of systematics if one injects with nonzero kappa_s, but ignores kappa_s in the recovery.

Abhirup asks if it makes sense to run the kappa test on the TGR relation injections, though since those are high-mass, perhaps create some for GW151226-like masses, since the infrastructure exists. Abhirup says that they can start by trying the test on the existing GW150914-like frames, since TIGER/FTA runs are being performed on those injections with dchi4, after Anuradha and Abhirup have finished going through the code. Krishnendu mentions that they did not obtain good bounds with GW150914, but Abhirup mentions that it’s still worth trying it on some non-GR injections.

Meet next week if there are things to discuss.

Abhirup will not be available for the two Fridays after next Friday (one due to the LVC meeting).