Rats! Another perplexing space mystery solved by science. New analysis of the famous “cold spot” in the cosmic microwave background reveals, and confirms, actually, that the spot is just an artifact of the statistical methods used to find it. That means there is no supervoid lurking in the CMB, and no parallel universe lying just beyond the edge of our own. What fun is that?

Back in 2004, astronomers studying data from the Wilkinson Microwave Anisotropy Probe (WMAP) found a region of the cosmic microwave background in the southern hemisphere in the direction of the constellation of Eridanus that was significantly colder than the rest by about 70 microkelvin. The probability of finding something like that was extremely low. If the Universe really is homogeneous and isotropic, then all points in space ought to experience the same physical development, and appear the same. This just wasn’t supposed to be there.

Some astronomers suggested the spot could be a supervoid, a remnant of an early phase transition in the universe. Others theorized it was a window into a parallel universe.

Well, it turns out, it wasn’t there.

Ray Zhang and Dragan Huterer at the University of Michigan in Ann Arbor say that the cold spot is simply an artifact of the statistical method–called Spherical Mexican Hat Wavelets–used to analyze the WMAP data. Use a different method of analysis and the cold spot disappears (or at least is no colder than expected).

');
}
//-->

“We trace this apparent discrepancy to the fact that WMAP cold spot’s temperature profile just happens to favor the particular profile given by the wavelet,” the duo says in their paper. “We find no compelling evidence for the anomalously cold spot in WMAP at scales between 2 and 8 degrees.”

This confirms another paper from 2008 also by Huterer along with colleague Kendrick Smith from the University of Cambridge who showed that the huge void could be considered as a statistical fluke because it had stars both in front of and behind it.

And in fact, one of the earlier papers suggesting the cold spot by Lawrence Rudnick from the University of Minnesota does indeed say that statistical uncertainties have not been accounted for.

Oh well. Now, on to the next cosmological mysteries like dark matter and dark energy!

By Nancy Atkinson
- Nancy Atkinson is currently Universe Today's Contributing Editor. Previously she served as UT's Senior Editor and lead writer, and has worked with Astronomy Cast and 365 Days of Astronomy. Nancy is also a NASA/JPL Solar System Ambassador.

So it begs the question: are there any other spots that were also in error. I find it hard to believe on a data set that large, there was only one location where the statistical package made a goof.

Cool!!!
I was impressed by the supervoid theory, but I never bought this “window to pararell universe”!
Anyway, science is what it is! And now we have to deal with reality, without magical gates to pararell universes!!!

Though I like the thoroughness of the approach the paper calls “superstatistic”, throwing both sets of data and sets of analysis methods on a peculiar observation.

“Spherical Mexican Hat Wavelets” – who says statisticians have no sense of humor…

It’s funny in the way of “spherical cows” – both are valid methods, until you start to milk them for specifics.

The kurtosis (non Gaussian distribution or deviation) in the CMB reported back in 2007 was thought to reflect some sort of “scar” in the early universe due to the interaction between two inflationary bubbles. I am personally not a big fan of some of these idea, so this result actually makes me rather happy.

So.. what if they see this “cold spot” because the method they first used looked for a certain type of data, and it found it, but their other methods look for other types of data. So what they’re actually seeing, is real, it’s just not exactly what they think it is.

Is there any point in even trying to explain what a spherical mexican hat wavelet is?

Is that a question?

I guess it is, since it ends with a question marks. So – yes. 😀

[I happen to have tried to use it once albeit only on 1D signals (but it turned out wasn’t useful for that particular application). So FWIW, if you meant you actually wanted the technical answer:

– A “wavelet” is the kernel of a wavelet transform, which looks like a wavelet, a localized wave.

It can be used as a filter to decompose a signal into a “wavelet power spectra”, i.e. tell you which components it contains. Or conversely, how to build a replica of the signal from your choose ‘palette’ of components.

Wavelets are useful for signals which are localized in space (and/or time) as the CMB fluctuations.

(As opposed to say fourier transform spectra which are more suitable for an ideal stationary signal, i.e. a repeating signal, than on a dynamic one.)

– A “mexican hat” function or wavelet looks like a crossection of one – a peak in the middle, swiftly going to zero after a “brim” dip. Here is a one-dimensional mexican hat. Apparently the technical name, as we are going technical, is Ricker wavelet.

– A spherical transform works in spherical coordinates, I assume. I.e. is isotropic, works the same, over the dimensions you are analyzing.]

So it begs the question: are there any other spots that were also in error. I find it hard to believe on a data set that large, there was only one location where the statistical package made a goof.

Wait, wasn’t this spot confirmed by radio telescopes too?

Cool!!!

I was impressed by the supervoid theory, but I never bought this “window to pararell universe”!

Anyway, science is what it is! And now we have to deal with reality, without magical gates to pararell universes!!!

Oh, I almost forget!

NOw we need to wait for some Plancks results and check out this area!

And I forgot to add:

“Spherical Mexican Hat Wavelets” – who says statisticians have no sense of humor… =-)

Rats indeed.

Though I like the thoroughness of the approach the paper calls “superstatistic”, throwing both sets of data and sets of analysis methods on a peculiar observation.

It’s funny in the way of “spherical cows” – both are valid methods, until you start to milk them for specifics.

There is something false about this so called fluke.

I do not think there ever was a fluke.

I think that there is a void.

Their explanation for this so called error gives me doubt and my intuition is right most of the time.

Find your own answers.Do not allow others to do all the thinking for you.

“Spherical Mexican Hat Wavelets” is ambiguous. Is it the Mexicans, the hats, or the wavelets that are spherical? Inquiring minds want to know!

“wiseguy Says:

September 16th, 2009 at 1:27 pm

“There is something false about this so called fluke.

I do not think there ever was a fluke.

I think that there is a void.”

Based on what exactly – your intuition?

Is there any point in even trying to explain what a spherical mexican hat wavelet is?

The kurtosis (non Gaussian distribution or deviation) in the CMB reported back in 2007 was thought to reflect some sort of “scar” in the early universe due to the interaction between two inflationary bubbles. I am personally not a big fan of some of these idea, so this result actually makes me rather happy.

LC

So.. what if they see this “cold spot” because the method they first used looked for a certain type of data, and it found it, but their other methods look for other types of data. So what they’re actually seeing, is real, it’s just not exactly what they think it is.

Is that a question?

I guess it is, since it ends with a question marks. So – yes. 😀

[I happen to have tried to use it once albeit only on 1D signals (but it turned out wasn’t useful for that particular application). So FWIW, if you meant you actually wanted the technical answer:

– A “wavelet” is the kernel of a wavelet transform, which looks like a wavelet, a localized wave.

It can be used as a filter to decompose a signal into a “wavelet power spectra”, i.e. tell you which components it contains. Or conversely, how to build a replica of the signal from your choose ‘palette’ of components.

Wavelets are useful for signals which are localized in space (and/or time) as the CMB fluctuations.

(As opposed to say fourier transform spectra which are more suitable for an ideal stationary signal, i.e. a repeating signal, than on a dynamic one.)

– A “mexican hat” function or wavelet looks like a crossection of one – a peak in the middle, swiftly going to zero after a “brim” dip. Here is a one-dimensional mexican hat. Apparently the technical name, as we are going technical, is

Ricker wavelet.– A spherical transform works in spherical coordinates, I assume. I.e. is isotropic, works the same, over the dimensions you are analyzing.]

“how to build a replica” – how to paint a replica, in the context.