"B_0^s" decays to "J/psi phi" and the Standard Model.
The 3.2-sigma statement is equivalent to a 99.9% confidence level while the 2.12-sigma statement is something like a 96% confidence level. But even for the higher confidence level, the proverb says that one half of the published 3-sigma results are wrong. It may actually be more than one-half and in less than ten days, we see some evidence of this general principle.
At a flavor physics and CP conference in Italy, the CDF Collaboration (see the detector on the picture) just announced the results of their own analysis of the "B_0^s" decays, using a significantly higher number of events than what they had analyzed together with D0 in the past. And the
Again, you can't really build science on 2-sigma (95%) and not even 3-sigma (99.7%) claims because most of them turn out to be wrong.
Thanks to Jester at Resonaances