my reading of that section is that they're just trying to illustrate what statistical error is. i don't think there's any real meaning to the 1000 value. [edi

my reading of that section is that they're just trying to illustrate what statistical error is. i don't think there's any real meaning to the 1000 value. [edit] but googling for "statistical significance planck anisotropy" turned up http://arxiv.org/abs/1303.5083 http://arxiv.org/abs/1303.5083 which says "3 sigma" which is about 0.3% (p=0.003). so it's close. [edit2] whoa. but you've got the right answer with the wrong maths. if you toss the same side of the coin 10 times in a row, that's p=0.001 (1/2^10 = 1/1024). getting the same side 1000 times in a row is much, much less likely. ps since the result is basically (if i understand right) "one side of the universe is slightly different from the other" then it depends on how you choose the halves. if you divided things "at right angles" to what gives the signal, you wouldn't see anything at all. not sure if that is what you meant by partitioning (mathematically they use something like a 3d fourier transform - spherical harmonics - and this is a significant deviation from zero for the lowest frequency component).