Fine Structure Constant Varies With Direction in Space

[sanman]

Here's something profoundly earth-shaking:

Thursday, August 26, 2010
http://www.technologyreview.com/blog/arxiv/25673/
A spatial variation in the fine structure constant has profound implications for cosmology

--

What is going to have to be revised as a result of this discovery? What are the likely implications for our understanding of the universe?

 

[Chronos]

Ahem, that is highly speculative - to be nice. Publication on arxiv lends little credibility. I've seen many such papers that are never cited by anyone other than the authors, and crackpot associates - and for good reasons [e.g., anomalous redshift].

 

[Vanadium 50]

This is by Webb and Flambaum (and collaborators) - the people who have been doing this for a while. They are not crackpots. However, they are probably not right either - their statistics is a bit dodgy, and I don't think it is nearly as significant as they think it is.

 

[turbo]

This is by Webb and Flambaum (and collaborators) - the people who have been doing this for a while. They are not crackpots. However, they are probably not right either - their statistics is a bit dodgy, and I don't think it is nearly as significant as they think it is.

They have been doing these observations and analyses for a very long time. They have pretty much defined the systematics that could have skewed results and defined the observations that might support or disprove their analyses. Good science, IMO. If theorists and nay-saying wannabes want to weigh in, they might want to start to construct some observational programs. Cosmology is an observational science. If observation does not accord with theory, then theory must be modified to accord with observation.

 

[Vanadium 50]

nay-saying wannabes

Wannabe? I have only one thing to say to that...<plbbbbt>

The "dodgy" statistics criticism comes about because of the well-known "look elsewhere effect" or sometimes called a trials factor. Specifically, one cannot properly assign a significance level to an observation when one has already selected the most significant discrepancy without incorporating the other places where one could have found an equally significant discrepancy.

As an example, if I do 100 tests and find one of them that has a 1% chance of being from chance alone, I cannot say that is statistically significant.

In this case, Webb et al. locate their dipole to +/- 9 degrees of declination and +/- 0.6 hours of RA. There are 200 such directions in the sky, so there is a trials factor of at least 200. It's higher still, since the bin center is chosen specifically to get the largest effect. That results in a minimum trials factor of 400. So their 4 sigma effect becomes a 2.4 sigma (or smaller) effect.

Furthermore, if you look at Figure 3, you can see that there is one point on the left, one cluster in the middle that can equally well fit a flat or sloped line, and one point on the right, arguably two. So the whole effect is driven by two points (arguably three). This is completely inconsistent with their Figure 4, where they additionally claim that by removing their most significant points, their significance actually goes up!

Hence, "dodgy".

 

[turbo]

The Wannabe was not aimed at you, Vanadium. You gave a reasoned response to what might be an insignificant statistical anisotropy.

 

[cristo]

The article in the OP is not published and, as per PF rules, is off topic for discussion. If anyone wants to cite an older, published article, that makes similar points, please PM me and I'll consider re-opening the thread.

Edit: A new thread has been started with suitable references. See here: https://www.physicsforums.com/showthread.php?t=425163