PureB

[yomori]

Implementation of hybrideb and NaMaster in pureB to compute B-modes.

[damonge]

Has anyone tested the standard B-mode purification method for cosmic shear data?
My expectation was always that it would not work at all, since it requires a smooth differentiable mask, which is very difficult to achieve for cosmic shear without losing most of the area.

[yomori]

My gut feeling is the same as yours (I think the apodization scale has to be pretty wide too), which is why I didn't even bother. Might be able to do measure something if we apply constrained inpainting or something first (at which point the measurement isn't too meaningful).

[xuod]

@damonge @yomori I did try pure B-mode on DES Y3, but because of all the tiny holes in the mask, the apodization drastically reduces the effective area (or equivalently reduces the weight of many areas in the footprint). So that makes the test lose statistical power.

[damonge]

Yep, indeed, this is the main problem with the standard purification approach. I could never convince myself that this would be useful for cosmic shear.

[yomori]

Yeah so thats why I have three modes of running this right now.
(a) Namaster with purify_b=False -- to replicate DES-Y3 what did.
(b) Namaster with purify_b=True -- if anyone wants to run it on idealized sim or something (just a flag change anyway).
(c) HybridEB -- the more robust way of computing BB.

If you are worried that someone will randomly run it with purify_b=True without knowing the caveats and use that in a paper, I can throw in a warning. If you really really want it to be taken out, I'm happy to do that too.

[joezuntz]

@yomori Do you remember where we were on this PR? If there's something fiddly that's not actually needed then we can merge it now and update later.