4 ms·

An interesting part from a different article mentioned here in the comments: The results do not meet the so-called five-sigma threshold of statistical certaint

An interesting part from a different article mentioned here in the comments: The results do not meet the so-called five-sigma threshold of statistical certainty that is the gold standard in physics for claiming a discovery. But many in the collaboration have shifted in recent months from a position of scepticism to confidently backing the finding. Source: https://www.theguardian.com/science/2025/mar/19/dark-energy-mysterious-cosmic-force-weakening-universe-expansion https://www.theguardian.com/science/2025/mar/19/dark-energy-...

sigmoid10 2y ago
If you combine it with other data it's from 2.8 to 4.2 sigma. Pretty strong and it has gone up since last year with more data. Nothing definite yet, but certainly intriguing.
- fedeb95 2y ago
  yes, that was my point! Interesting that's "on the rise"
mr_mitm 2y ago
I think it's worth noting that the five sigma standard traditionally comes from particle physics, though. Cosmology doesn't have the luxury of being able to generate arbitrary amounts of data, and in fact has been a precise science for only sixty years or so. Hubble's first measurement of the expansion of the universe was off by a order of magnitude.
rcxdude 2y ago
That's basically a particle physics threshold, because in particle physics you tend to do many, many experiments (usually you're doing thousands a second). It's hard to get something that certain in astronomy, where you've only got one universe to observe.
- gpderetta 2y ago
  It is also an informed threshold. FTL neutrinos where, IIRC, above 6 sigma and nobody believed it was true, while there were strong rumors that Higs had been found well before it hit 5 sigmas.
  - rcxdude 2y ago
    Yeah, statistical significance can tell you nothing about systematic errors you're unaware of or errors in the statistics itself
antognini 2y ago
As a former astronomer, one thing to highlight here is that statistical uncertainties aren't usually all that important in assessing how likely a result in astronomy is to be true. The main issue you grapple with in the field are your systematic uncertainties. When you report your uncertainty, it always is an uncertainty in the context of some model. However, if certain assumptions that your models have made are wrong, you will have underestimated your true uncertainty. Unfortunately, systematic uncertainties can't be reduced just by getting more data --- you have to somehow verify that the different assumptions your model has made are correct. Astronomers spend a lot of time arguing with each other about whether they have properly incorporated all their systematic uncertainties. For measurements of dark energy this very quickly gets you into the weeds of Type Ia supernova physics (which is made more difficult because we don't know for certain what Type Ia supernovae are), stellar physics in Cepheids, the effect of metallicity, selection bias effects, and on and on.
- fedeb95 2y ago
  interesting, thanks. I can't edit my old comment, but my original intention wasn't to discredit the news, or the study. I will word my comments better in the future.

sigmoid10 2y ago

If you combine it with other data it's from 2.8 to 4.2 sigma. Pretty strong and it has gone up since last year with more data. Nothing definite yet, but certainly intriguing.

fedeb95 2y ago
yes, that was my point! Interesting that's "on the rise"

mr_mitm 2y ago

I think it's worth noting that the five sigma standard traditionally comes from particle physics, though. Cosmology doesn't have the luxury of being able to generate arbitrary amounts of data, and in fact has been a precise science for only sixty years or so. Hubble's first measurement of the expansion of the universe was off by a order of magnitude.

rcxdude 2y ago

That's basically a particle physics threshold, because in particle physics you tend to do many, many experiments (usually you're doing thousands a second). It's hard to get something that certain in astronomy, where you've only got one universe to observe.

gpderetta 2y ago
It is also an informed threshold. FTL neutrinos where, IIRC, above 6 sigma and nobody believed it was true, while there were strong rumors that Higs had been found well before it hit 5 sigmas.
- rcxdude 2y ago
  Yeah, statistical significance can tell you nothing about systematic errors you're unaware of or errors in the statistics itself

antognini 2y ago

As a former astronomer, one thing to highlight here is that statistical uncertainties aren't usually all that important in assessing how likely a result in astronomy is to be true. The main issue you grapple with in the field are your systematic uncertainties. When you report your uncertainty, it always is an uncertainty in the context of some model. However, if certain assumptions that your models have made are wrong, you will have underestimated your true uncertainty. Unfortunately, systematic uncertainties can't be reduced just by getting more data --- you have to somehow verify that the different assumptions your model has made are correct. Astronomers spend a lot of time arguing with each other about whether they have properly incorporated all their systematic uncertainties. For measurements of dark energy this very quickly gets you into the weeds of Type Ia supernova physics (which is made more difficult because we don't know for certain what Type Ia supernovae are), stellar physics in Cepheids, the effect of metallicity, selection bias effects, and on and on.

fedeb95 2y ago
interesting, thanks. I can't edit my old comment, but my original intention wasn't to discredit the news, or the study. I will word my comments better in the future.