Weakening evidence for dark energy

[Eebster the Great]

A friend of mine linked me an article titled "Marginal evidence for cosmic acceleration from Type Ia supernovae" which claims that a recent survey of a large sample of Type Ia supernovae is consistent with a constant rate of expansion (p > 0.003). This is surprising, given a five sigma certainty was claimed in Nobel prize-winning research in the 1990s, which essentially forms the basis of our understanding of dark energy. Granted, the evidence still supports an accelerating expansion more strongly than a constant expansion, but the idea that it may be consistent with either is highly significant, and could mean we should reconsider how sure we are that dark energy is real at all.

I don't understand the statistical techniques used in the study and don't know how much weight to give this article. Does anyone think this is overblown, or is it a big deal?

For easier reading, here is an article from space.com on the study.

[DaBigCheez]

The evidence for accelerating expansion in this study being "only" three sigma rather than five doesn't sound to me like a compelling evidence to say "it doesn't exist". I'm not sure if you can really stack standard deviations in this manner, but wouldn't a 95% CI around a "true" value that's a five-sigma result encompass a three-sigma result?

[Weeks]

I don't understand the statistical techniques used in the study and don't know how much weight to give this article. Does anyone think this is overblown, or is it a big deal?

SciAm says to disregard.

[Eebster the Great]

The evidence for accelerating expansion in this study being "only" three sigma rather than five doesn't sound to me like a compelling evidence to say "it doesn't exist". I'm not sure if you can really stack standard deviations in this manner, but wouldn't a 95% CI around a "true" value that's a five-sigma result encompass a three-sigma result?

I'm not sure I understand the question. The idea however is that if the results from supernova surveys become increasingly consistent with constant expansion as the samples get larger, then initial data could have been due to sampling bias. Three sigma is not considered significant in astronomy in most cases, even though it is "only" a 0.3% probability.

I don't understand the statistical techniques used in the study and don't know how much weight to give this article. Does anyone think this is overblown, or is it a big deal?

SciAm says to disregard.

Thanks, that's exactly what I was looking for, especially:

[The authors] assume that the mean properties of supernovae from each of the samples used to measure the expansion history are the same, even though they have been shown to be different and past analyses have accounted for these differences.

and

The present study also ignores the presence of a substantial amount of matter in the Universe, confirmed numerous times and ways since the 1970’s, further reducing the study confidence.

What specifically is that second quote referring to?

[gmalivuk]

The evidence for accelerating expansion in this study being "only" three sigma rather than five doesn't sound to me like a compelling evidence to say "it doesn't exist". I'm not sure if you can really stack standard deviations in this manner, but wouldn't a 95% CI around a "true" value that's a five-sigma result encompass a three-sigma result?

A 95% CI is 1.96 sigma, and also that assumes you're talking about the same sigma in both cases, which you probably aren't.

Probabilistically, the chance of being outside 5 sigma is 1 in 1.7 million, and the chance of being outside 3 sigma is 1 in 370. I think Bayes theorem involves simply multiplying odds ratios (if I remember correctly), so given prior 1700000:1 odds against not-accelerating, 370:1 odds for not-accelerating is 1:370 odds "against" not-accelerating, for a result of 1700000:370, or about 4600:1.

[TvT Rivals]

(No time now for reading the paper, but:)

So the universe doesn't have to die in a Big Rip? Hooray!

[doogly]

Double false! The claims in the paper were not so legit, but we aren't likely to head for a rip even with the currently understood acceleration.

[Frenetic Pony]

Double false! The claims in the paper were not so legit, but we aren't likely to head for a rip even with the currently understood acceleration.

More like overblown (to get attention?). It's not like the claims are mathematically incorrect, just that a larger survey of one part of the evidence for dark energy is not as definitive as previous measurements and may even fall below the arbitrary "statistical significance" point of 5 sigma.

Can't discount the "possibility" of dark energy being a statistical anomaly, but an outside chance of one type of measurement not holding the theory up to high scrutiny is hardly convincing enough, by itself, to be really significant.

[Xanthir]

Double false! The claims in the paper were not so legit, but we aren't likely to head for a rip even with the currently understood acceleration.

IIRC, we're currently theorized to be headed for the Big Freeze (infinite expansion, but not at an infinitely accelerating rate), right?

[doogly]

Word.

[jewish_scientist]

If our understanding of supernovas was flawed in some way, then could this new data be consistent with the theory of dark energy? In other words, can this new data be interpreted as evidence that our theories associated with supernovas are incorrect?

NOTE: I am just curios if it is possible to interpret the study this way, not that we should.

[Mike Rosoft]

Just a few days ago I watched a video explaining the result. (The bottom line is: yes, the observation of supernovae is by itself consistent with no dark energy/no acceleration of expansion; but combined with the other evidence - 1) the amount of observed matter in the universe and 2) the observation that the universe is relatively flat - we are pretty certain [well over 5-sigma] that dark energy is still a thing.)

[thoughtfully]

I'm not sure if another citation is needed, but Sean Carroll at Caltech writes an awesome blog and this is his field (cosmology)
https://www.preposterousuniverse.com/bl ... rk-energy/