3 ms·

My impression (from well off to the side in a tangentially related field, mind you) has been that false positives are common enough in the individual detectors

by Steuard 11y ago

My impression (from well off to the side in a tangentially related field, mind you) has been that false positives are common enough in the individual detectors that they really do need the comparison with the other location's data as a filter. (Not common enough for an actual coincidence to be at all likely, though.) I'm happy to be corrected if I'm wrong, of course! But I think the system has only been designed to be reliable as a whole.

lutusp 11y ago
> But I think the system has only been designed to be reliable as a whole. Yes, absolutely, I only meant that initial evaluations might be more objective if the two observatories cannot compare notes (at first) -- that might lead to a kind of confirmation bias. Only after one observatory satisfies itself that they have a sufficiently unambiguous signal should they be allowed to correlate with the other observatory's results. Considering the absence of plausible signals to date, this idea might be bending over backwards to avoid observational bias.
- Steuard 11y ago
  I see your point, but I still have a sense that the false positive rate is just too high. I don't know the actual false positive rate, but I've somehow pictured maybe a dozen or so very brief (<1 sec?) signals per year. Under anything like those circumstances, an accidental coincidence would still be vanishingly unlikely (since there are 3x10^7 sec/year, it would be something like a one in a million chance), but the collaboration can't be expected to do full, detailed analysis of false signals a dozen times a year.
  - lutusp 11y ago
    > ... I still have a sense that the false positive rate is just too high. That might be true in principle, but considering what's at stake, I think the investigators will bend over backward to avoid making a claim that isn't ironclad. Remember that in real science, investigators look for explanations that will falsify their results, and only give credence to results that resist sincere efforts at falsification. (Contract this to the outlook of a pseudoscientist, who only looks for reasons to confirm his results.) Remember the incident in which some investigators claimed that neutrinos were exceeding the speed of light? In truth, they were looking for reasons to confirm their result, not falsify it. They were wrong, the real cause was a bad connection and ... they were pressured to resign their positions. https://en.wikipedia.org/wiki/Faster-than-light_neutrino_anomaly https://en.wikipedia.org/wiki/Faster-than-light_neutrino_ano... > ... the collaboration can't be expected to do full, detailed analysis of false signals a dozen times a year. That's true, but they could instead choose a handful of genuinely peculiar signals, signals unlikely to have natural causes, and try to falsify those. As above, there are a handful of cosmological sources that would produce -- not a brief signal -- but a pattern of signals, like a rising tone created by merging black holes. Those would be more interesting, as well as being easier to compare to false terrestrial sources.

by Steuard 11y ago

lutusp 11y ago

> But I think the system has only been designed to be reliable as a whole. Yes, absolutely, I only meant that initial evaluations might be more objective if the two observatories cannot compare notes (at first) -- that might lead to a kind of confirmation bias. Only after one observatory satisfies itself that they have a sufficiently unambiguous signal should they be allowed to correlate with the other observatory's results. Considering the absence of plausible signals to date, this idea might be bending over backwards to avoid observational bias.

Steuard 11y ago
I see your point, but I still have a sense that the false positive rate is just too high. I don't know the actual false positive rate, but I've somehow pictured maybe a dozen or so very brief (<1 sec?) signals per year. Under anything like those circumstances, an accidental coincidence would still be vanishingly unlikely (since there are 3x10^7 sec/year, it would be something like a one in a million chance), but the collaboration can't be expected to do full, detailed analysis of false signals a dozen times a year.
- lutusp 11y ago
  > ... I still have a sense that the false positive rate is just too high. That might be true in principle, but considering what's at stake, I think the investigators will bend over backward to avoid making a claim that isn't ironclad. Remember that in real science, investigators look for explanations that will falsify their results, and only give credence to results that resist sincere efforts at falsification. (Contract this to the outlook of a pseudoscientist, who only looks for reasons to confirm his results.) Remember the incident in which some investigators claimed that neutrinos were exceeding the speed of light? In truth, they were looking for reasons to confirm their result, not falsify it. They were wrong, the real cause was a bad connection and ... they were pressured to resign their positions. https://en.wikipedia.org/wiki/Faster-than-light_neutrino_anomaly https://en.wikipedia.org/wiki/Faster-than-light_neutrino_ano... > ... the collaboration can't be expected to do full, detailed analysis of false signals a dozen times a year. That's true, but they could instead choose a handful of genuinely peculiar signals, signals unlikely to have natural causes, and try to falsify those. As above, there are a handful of cosmological sources that would produce -- not a brief signal -- but a pattern of signals, like a rising tone created by merging black holes. Those would be more interesting, as well as being easier to compare to false terrestrial sources.