That's going to be a figure in my thesis, but I can get to it sooner rather than later. -Chris On Fri, Apr 26, 2013 at 12:27 PM, John Kovac &lt;jmkovac(a)cfa.harvard.edu&gt;wrote;wrote: > > Hi Chao-Lin and Clem, > > > On 4/25/13 6:23 PM, Clem Pryke wrote: > >> there is a large increase in variance when rel gain deproj is >>> turned on. It is somewhat a surprise because there shouldn't be >>> too many modes removed, naively. >>> >> >> We have known this for some time and I am interested to hear that you >> think it's surprising. >> >> Chris investigated the E/B mix which comes from deproj in this post: >> >> http://bicep.caltech.edu/~**spuder/analysis_logbook/** >> analysis/20120925_sim1227/aps_**pager.html<http://bicep.caltech.edu/~spu… >> >> Click: >> >> - yes pol (no systematics) >> - n2048/n512 no common A/B >> - Kendrick >> >> We are seeing the residual E/B mixing which comes from our dumb ra/dec >> flat sky projection. Click the "n2048... arc proj button" to see this >> reduced further. >> >> Going back to the "n2048/n512 no common A/B" option we can now click >> deproj="relgain" and see a large increase in the E/B mix - and scatter >> thereof. >> >> This is why I have been been keen not to use relgain if we can possibly >> help it. >> >> Now click "n2048/n512 ... noisy template" - not much change. >> >> Now click deproj type="per-scanset" - this is fitting the deproj >> coefficient per 50 minute scanset rather than per phase (10 scansets). We >> see increase in the amplitude - and scatter - of the E/B mixing. >> >> So this last matches your expectation. >> > > Thanks for the reminder of these links--of course we've been studying this > increased variance for a long time, including the first deprojection work > Chris did way back in 2009(?) and Randol's work on BICEP1. > > I for one was surprised to be reminded, clicking through this pager, that > the effect of the arc vs RAdec projection is subdominant to relgain deproj, > but not to diffpoint deproj. Is that right? If so, and if you've been > hoping we can avoid relgain deproj, then why haven't we been pushing the > arc projection? > > > > > >> I suspect that as our integration time increases >>> >> >> However I suspect the leakage does not go up with integration time - if >> it did it would be way more in Chris sim1234 which is full tag set sims >> versus the pager above which is tag subset. >> >> and also as our number of pairs increases (from B2 to Keck), the # of >>> modes we remove also increases.. >>> >> >> From Stefan's plots I think it is not vastly larger. We should check... >>> >> > I think the right way to think about this is loss of modes per pairmap. > As you coadd more pairmaps, including from different detectors, the loss > of information doesn't actually get worse. > > But I agree we should test this quantitatively using the sims we've got > now! > > John > > > -- > ______________________________**______________________________**_______ > John Kovac jmkovac(a)cfa.harvard.edu > > Assistant Professor, Astronomy and Physics, Harvard University > 160 Concord Ave rm 310, Cambridge MA 02138, 617-496-0611 > > ______________________________**_________________ > KeckArray mailing list > KeckArray(a)lists.stanford.edu > https://mailman.stanford.edu/**mailman/listinfo/keckarray<https://mailma… > -- ********************************************************************** Christopher Sheehy - Ph.D. candidate - University of Chicago Research Specialist, University of Minnesota, Department of Physics Room 220 Tate, 116 Church Street S.E. Minneapolis, MN, 55455 Tel: 612-625-1802 Fax: 612-624-4578 email: csheehy(a)uchicago.edu **********************************************************************