https://world.hey.com/dhh/basecamp-s-new-etiquette-regarding...

Naturally, much more statistical analysis has been done to ground the claims of "detection"; beyond detailed academic publications, LIGO and others have been producing layperson-accessible science summaries for years/decades that address these and other questions.

Citation please. Every layperson accessible summary has said "we use advanced statistics and machine learning" and I haven't found a simple high school statistics accessible explanation yet. Unlike say the higgs boson, I think for this experiment a simple statistical treatment is not an unreasonable request.

Please show me and correct me. I would love to be able to believe we have detected gravitational waves.

Probably not high school statistics accessible but Bayesian statistics and Gaussian noise and GR isn't.

> This article was originally published at Aeon and has been republished under Creative Commons.

An awful lot of evil is done in this world in the name of doing what's best for others against their will.

The instrument data of this event is also available to the public at https://losc.ligo.org/events/GW170104/

>"GW170104 was first identified by inspection of low-latency triggers from Livingston data [15–17]. An automated notification was not generated as the Hanford detector’s calibration state was temporarily set incorrectly in the low-latency system. After it was manually determined that the calibration of both detectors was in a nominal state, an alert with an initial source localization [18,19] was distributed to collaborating astronomers [20] for the purpose of searching for a transient counterpart. About 30 groups of observers covered the parts of the sky localization using ground- and space-based instruments, spanning from γ ray to radio frequencies as well as high energy neutrinos [21]." https://dcc.ligo.org/LIGO-P170104/public

Regarding the earlier detection:

>"At 11:23:20 UTC, an analyst follow-up determined which auxiliary channels were associated with iDQ’s decision. It became clear that these were un-calibrated versions of h(t) which had not been flagged as “unsafe” and were only added to the set of available low latency channels after the start of ER8. Based on the safety of the channels, the Data Quality Veto label was removed within 2.5 hours and analyses proceeded after re- starting by hand." http://ligo.elte.hu/magazine/LIGO-magazine-issue-8.pdf

So both times humans had to take special action for the detection to "count". I really wonder about whether the null model they are using is appropriate/relevant here.

Also, the other thing I have been concerned about is the lack of any corroborating evidence that these signals are truly generated by inspiraling black holes(gamma ray bursts, etc). Apparently, in this case the above-mentioned miscalibration has impeded that effort:

>"The event candidate was not reported by the low-latency analysis pipelines because re-tuning the calibration of the LIGO Hanford detector is not yet complete after the holiday shutdown. This resulted in a delay of over 4 hours before the candidate could be fully examined. We are confident that this is a highly significant event candidate, but the calibration issue may be affecting the initial sky maps. We will provide an update in approximately 48 hours which may include an improved sky map." https://gcn.gsfc.nasa.gov/other/G268556.gcn3

I can't tell from that text file whether they got corroborating evidence or not. IANAP though.

By design, detection statements and significance estimates come solely from the offline analysis which is conducted separately (i.e. not triggered by) the online analysis. No human intervention is required here, as the issue with the online status information was known about at the time and was not an issue with the data itself. Even if there were no candidate events at the time, it would be been included in the offline analysis of the period containing the event.

In regards to GW150914 and iDQ, you should know that iDQ has never been approved as a veto for CBC (compact binaries such as neutron stars and black holes) searches. Again, no intervention was required to "remove" the veto as it was never used in the offline analysis nor would be in the first place. It's only use that I am aware of is as a veto against Burst triggers in online analysis. These searches look for generic signals, but may also detect some of the louder CBC sources, such as GW150914. In case you were wondering, there weren't dedicated online CBC searches at the time of GW150914, but there were offline analysis, and those produced the results reported in the original detection paper.

This is information that should be included in the papers because the current description is too terse. Basically you are saying that the filters used for online analysis have nothing to do with the background model, zero influence on what periods get included, etc. I'm still unclear on what exactly needed to be "restarted by hand" for the original GW150914 signal, but ok.

>"In the case of binary black holes, the expectation is that there will be no electromagnetic counterpart, as the region is expected to be cleared of matter well before we observed the black holes merging."

Is there any other type of event that is expected to be accompanied by some kind of corroborating evidence?

In terms of corroborating evidence, remember that the two independent LIGO detectors - 3000km apart - saw the event within 10ms of each other. That's enough corroborating evidence for a lot of people. The NASA text file you link shows no observed electromagnetic counterpart, but that's expected: unfortunately the best models so far for black hole coalescences predict very little or no electromagnetic emission - so although EM partners were informed, the chances of them seeing anything were slim. Other predicted sources of gravitational waves, like as-yet unseen binary neutron star coalescences, are more likely to emit EM radiation and stand a chance of being witnessed by conventional observatories as "corroboration".

I don't see what the first part of the post has to do with the null ("background noise") model being inapplicable to situations where special human intervention comes into play. Do they include any events like that in the background timeseries or not? I am suspecting not (which renders the model false and hence false alarm rates/sigma values meaningless), but do not know for sure.

Second, that is just a detection. Corroboration occurs when your model predicts multiple types of observations related to a phenomena (measured by different types of instruments). This weakening of definitions is concerning to me if it has infected physics. I have seen that trick used a lot by "softer" fields such as medicine/psych (eg their definition of a replication is just seeing "an effect" in the same direction).

I guess we will see once the count gets into the dozens. If it happens without any kind of outside way to verify these signals are inspiraling black hole events it will definitely be interesting to see how the physics community deals with it.

On the instrumental side, we've only just reached the sensitivity levels to make any detections in the first place, so it's not surprising that we're not getting a huge number of events (otherwise the previous generation of detectors would've seen something). In addition, each individual observatory has its own "antenna pattern", making us less less sensitive to certain sky locations. This will improve as VIRGO, KAGRA, and LIGO-India come online in the future.