TGR Review
-
"For pipelines that are not testing a specific alternative theory or specific physical scenario [BURST ANALYSIS], they should be demonstrated to be able to pick up a GR deviation for some non-GR waveform that is not the one used in the analysis [BW RECONSTRUCTION]. One can use the waveforms from the non-GR MDC, but should not feel required to use these if another non-GR waveform is more appropriate."
-
This feels important: The pipeline should be run on a sufficiently wide variety of GR injections that one can be sure that there won't be any problems sampling the posterior with the settings used or issues with the GR deviations causing problems for the data analysis, as occurred for the MDR test, where large amounts of dispersion made the waveform extend beyond the analysis segment and thus wrap around. Thus, it is likely necessary to consider both high- and low-SNR injections, since the low-SNR case is likely where issues with more extreme GR deviations will show up.
- Test on low and high SNR injections
- Create XML using LAL CLI link, read in samples (or write code to parse XML directly), and pass to iwc_pipe for better SNR gauging (?)... just an idea, here.
- The pipeline should return GR for (quasicircular) GR BBH injections including precession and the full mode content available, ideally an NR injection. (We should have some BBH NR injections for the systematics MDC, but these are not yet created.)
- Make sure that NR waveforms go nicely with SimInspiral I/O package, and that we know how to do this, if we need to do this (should be fine; SimInspiral I/O stuff entirely gotten from python-ligo-lw and GstLAL code). It seems as if this is just asking for the return of the GR waveform... but we are using a Burst analysis, so feels as if this is satisfied.
-
- Pipelines that are applicable to BNS and NSBH signals should be run on at least a few representative GR injections of these systems, ideally NR (hybrid) injections using a different waveform model to make the hybrid than the model used in the test, but at least using injections of different model waveforms than the ones used in the test.
- "different model waveform", again, seems satisfied by our use of Burst modeling technique. Making sure that we can do NR-hybrid injections seems most pertinent (?)
- The pipeline should be run on at least one real event. For pipelines that have run in the past and are being upgraded, they should compare with previous results on a given event.
- Base-functionality. Almost done.
- For tests that make statements about compact binary physics, not just deviations in waveforms, verify that all systematic errors are understood and sufficiently well under control in order to make whatever statements are proposed for the paper.
- Ours is modeling excess power, should have no systematics to worry about.
- The review should assess that the selection criteria proposed are clearly stated, appropriate, and justified.
- Working on that here...
- The reviewers should also provide instructions on a wiki page for running the test on an event (e.g., how to set up and submit a Condor job to the cluster for most tests), and these should be reviewed for accuracy and clarity.
- Done. Can clean up, some more, for Katerina's replacement.
- The review statement should mention the analysis's selection criteria and proposed contributions to the O4 TGR papers (i.e., the specific outputs that will appear in the papers, such as constraints on parameters, Bayes factors, etc.). It should state any potential issues with the test and its interpretation (e.g., difficult corner cases that could arise), and also discuss any potential sources of systematics and how they are addressed/mitigated, if applicable. The statement should also give a list of references for the test (i.e., the papers that should be cited for the test in the O4 TGR papers) and also give the computational cost (SUs or at least CPU hours) and memory usage for the GW150914-like and GW170608-like GR injections (as appropriate) from the systematics MDC.
- Going to let others chime in here on literature (GW150914 TGR paper, O1 TGR paper (some theory specifics), O3b TGR paper)e
- We don't have MDC's done, but maybe we can minimally by review time in order to generate this basic information.