Another brief paper on astro ph today looking at the fine structure constant: Rahmani et al. Instead of using the many multiplet method of Dzuba et al & Webb et al—which (if there are no systematic errors) can estimate da/a down to one part in 100,000 in a single quasar (absorption) spectrum—the authors of this study have used the older method of Bahcall et al—which can only estimate da/a to about one part in 10,000 in a single spectrum (but which should be less prone to systematic errors). The point being that if indeed there are no systematic errors (in either case) then given enough observations you’ll be able to measure the mean da/a at a given pointing and/or redshift down to arbitrarily small accuracy. In this study Rahmani et al have about 2500 quasars and get down to the (claimed) accuracy level of approximately a single many multiplet method estimate. Interestingly, the authors find a slope of da/a vs loopback time (in Gyr) consistent with zero: -(0.9 +/- 1.1) x 10^-5.
Obviously, if you believe that the many multiplet method has absolutely no systematic error then Rahmani et al are just wasting their time, and Webb & King should rock up for their Nobel prize right away, since there’s obviously 4sigma evidence for a fine structure dipole already …
I also noticed that the Allison & Dunkey paper with the incorrect nested sampling algorithm has been accepted by MNRAS and is available in early/online form. I’m a bit surprised that although I wrote to them to explain their mistake* and blogged about it here they chose not to correct their manuscript?! It’s quite obviously wrong and one can verify its wrongness through numerical experiment if you’re not convinced by theory.
* the mistake: The algorithm given in their 5 point description of nested sampling is only valid for the case of uniform priors! Ellipse-based nested sampling is fine, but you have to specify this key constraint!!