Possible error found in p_astro for GWTC-4.0: Candidate data release

Data Analysis
1

In the latest release data (published May 22, 2026) from GWTC-4.0: Candidate data release , in the p_astro data, event GW230529_181500 has the following probability distribution.

Screenshot 2026-06-10 at 6.24.47 PM

The candidate release p_astro lists that the event has highest likelihood of being BNS, however according to this paper the event is most likely an NSBH.

I am not sure how to interpret this. If possible, please check if this is an error and if p_astro is impacted for other events. Thank you!

1 Like
2

Hi, I looked into this a bit using the data from the Zenodo link you quoted (GWTC-4.0: Candidate data release) and the /GWTC4p0_search_data_release.ipynb Jupyter notebook included with that release.

The p_astro data listed in your message above are the results from the MBTA pipeline. We can verify this, by adding a cell with this content to the sample Jupyter notebook.

with open(pathToData+'mbta/pastro/IGWN-GWTC4p0-38214bd95_724-GW230529_181500-mbta_SourceClassification_SearchResultDataRelease.json', 'r') as fp:
    pastro_MBTA_GW230529_181500 = json.load(fp)
pastro_MBTA_GW230529_181500

We get the output:

{'Terrestrial': 0.00011697,
 'BNS': 0.790605186,
 'BBH': 0.0,
 'NSBH': 0.209277843}

However, as mentioned both in the discovery paper and in the GWTC-4.0 catalog paper, GW230529_181500 was found by 3 pipelines in total:

  • MBTA
  • PyCBC
  • and GstLaL

Adding similar cells for the other two pipelines we get p_astro results consistent with the NSBH scenario.

For PyCBC:

Adding the following cell to the jupyter notebook

with open(pathToData+'pycbc/pastro/IGWN-GWTC4p0-38214bd95_724-GW230529_181500-pycbc_SourceClassification_SearchResultDataRelease.json', 'r') as fp:
    pastro_PyCBC_GW230529_181500 = json.load(fp)
pastro_PyCBC_GW230529_181500

gives this output:

{'Astro': 0.9999763667039779,
 'Terrestrial': 2.3633296022113376e-05,
 'BNS': 0.3336530202820449,
 'NSBH': 0.666323346421933,
 'BBH': 0.0}

For GstLAL

Adding the following cell to the jupyter notebook

with open(pathToData+'gstlal/pastro/IGWN-GWTC4p0-38214bd95_724-GW230529_181500-gstlal_SourceClassification_SearchResultDataRelease.json', 'r') as fp:
    pastro_GSTLAL_GW230529_181500 = json.load(fp)
pastro_GSTLAL_GW230529_181500

gives this output:

{'BBH': 0.0,
 'BNS': 0.0,
 'NSBH': 0.852639814103928,
 'Terrestrial': 0.14736018589607205}

Different pipelines make different choices internally regarding some specifics of the p_astro calculation (e.g. the construction of the foreground (signal) and background (noise) distributions or how templates are split into bins in the chirp mass and mass ratio parameter space, etc). Therefore, the same event may be categorised as being more likely a NSBH by some pipelines or a BNS by others.

You can find more details, e.g in Section 3 of the latest GWTC-5.0 methods paper and references therein. Additionally, Appendix D.7 of the GWTC-3.0 catalog paper might also be useful.

Let me also stress that recently the LVK collaboration released GWTC-5.0, the 5th version of the Gravitational Wave Transients Catalog. You can find more information and links to the Data Products and Publications in the respective GWOSC GWTC-5.0 documentation page.

1 Like