I've put up a brief posting here:
http://bicep0.caltech.edu/~spuder/analysis_logbook/analysis/20131113_signfl…
pointing out that our sign-flip noise realizations are not assigning
the same sign to simultaneous BICEP2/Keck Array tags, and therefore are
not simulating noise correlations between the experiments. This was
surprising to me, as I thought the major benefit of sign-flipping noise
sims was the ability to account for correlated noise.
Our sims then have zero noise correlation between BICEP2 and Keck
Array, and we're not applying any noise debias to the cross-spectra.
Perhaps all of this is OK if we can prove that the noise correlations
are negligible, but I don't recall that having been demonstrated. Has
there been a deliberate decision not to match the sign-flip sequences,
or does it indicate something broken in the driver code?
- Walt