3
$\begingroup$

I am not sure how to read the numbers in this excerpt from the abstract of GW151226: Observation of Gravitational Waves from a 22-Solar-Mass Binary Black Hole Coalescence:

enter image description here

Are the +/- small numbers maximum observed deviations? Or $1\sigma$ or $5\sigma$ deviations? Or something else?

$\endgroup$
9
$\begingroup$

You cut off the sentence that tells you what the numbers mean. "All uncertainties define a 90% credible interval". Crudely speaking, it means that there is a 90% probability of the parameters lying in the quoted range, with the most likely estimate being the headline number.

It doesn't really make sense to translate these into Gaussian sigmas (it would be about 1.6 sigma), because the distributions are clearly asymmetric (judging by the asymmetric error bars).

$\endgroup$
  • $\begingroup$ Thanks! Is this standard practice in papers these days, or are there different ways to report numbers? $\endgroup$ – Frank Jun 16 '16 at 13:19
  • $\begingroup$ There are many ways to report measurements with uncertainty. The important thing is that one states the method being used, as these authors did. $\endgroup$ – Paul T. Jun 16 '16 at 13:35
  • $\begingroup$ Thanks Paul - I was just curious, as one could imagine that e.g. PRL mandates a certain format to ensure uniformity across the publications. $\endgroup$ – Frank Jun 16 '16 at 14:06
  • $\begingroup$ @Frank It really is up to the authors. What I would say is that if you know that your measurement uncertainty has non-Gaussian tails then it would be very misleading to only quote a "1 sigma" uncertainty. $\endgroup$ – Rob Jeffries Jun 16 '16 at 14:43
  • $\begingroup$ Yes indeed. But if the distribution is not gaussian, isn't it better then to show the whole distribution? There could be modes, fat tails, ... $\endgroup$ – Frank Jun 16 '16 at 14:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.