Looking at the pdf of the paper linked at slashdot, it seems the author is making a fairly weak claim. He found a direction that maximized the correlation of the elliptical axes with that line of sight, and reports the finding that the correlation is great enough to be extremely unlikely to be due to chance. However, the degree of correlation is small. The average ellipticity is 0.225 and he “bins” the data by the sine of the angle with the assumed preferred axis and finds that the best fit is a line with average ellipticity of 0.220 looking along the axis, and 0.229 looking perpendicular to it.
If the galaxies all had ellipticities of 0.225, and were all aligned with the axis, the average ellipticity would be 0 looking along the axis.
Note also that his worst fit finds 0.003 of the ellipticity accounted for by correlation, as opposed to 0.009 in the best fit.
With a lot of data, the chi-squared method very commonly gives a low “probability of rejecting the null hypothesis” along with a weak correlation, and textbooks warn of this. Note that the method merely asserts that there is SOME source of correlation other than chance, and this could easily be in the data collection process itself.
I was just about to say that very thing. Really...
I don’t know anything about this topic but I do wonder about magnetism. I wonder if we are clueless about its significance.
Maybe, but consider two things. First there's evidently some anistropy in the WMAP that seems to match up. Second, the fellow mentions another analysis he did with some tens of thousands of spriral galaxies in the same database. It's a lot fewer than the 200K elipticals because he needed to be able to resolve the direction of rotation. Again he found a preferred direction. So that's three independent lines of evidence. I'd say something's up.