8 Feb
2013
8 Feb
'13
4:48 a.m.
Hi Walt,
This is very interesting stuff and the kind of free thinking we are very much
in need of.
I agree with John that sim comparison is required - if I am understanding
correctly the two halves of the split are differently filtered versions of the
sky and hence not expected to match. I find it hard to guess the degree of
mismatch expected.
It always starts at the same phase, so it will be in
phase from one
scanset to another, and will add up over tags rather than averaging down.
It dies down over tens of minutes, so it does not form a fixed pattern in
all the half-scans of a scanset. This is meant as an illustration of the
type of signal that would not be effectively removed by our current
filtering and ground subtraction, and would't be caught by our
existing jackknife tests.
I agree it wouldn't be removed by ground subtraction. However I think it would
fail deck jack unless it was the same in every pair. And if it was the same in
each pair the false signal would be nearly the same row-to-row in the maps. (A
pair specific signal would fail deck jack unless each pair somehow "knew" to
adopt the same false signal as the pair previously at that elevation when the
array was rotated.)
Clem
--
**********************************************************************
Clem Pryke - Associate Professor - Physics
University of Minnesota,
Room 313 Tate, 116 Church Street S.E. Minneapolis, MN, 55455
Tel: 612-624-7578 Fax: 612-624-4578 email: pryke(a)physics.umn.edu
**********************************************************************
8 Feb
8 Feb
10:04 a.m.
New subject: Analysis note: half-scan jackknife and possible culprit
if I am understanding
correctly the two halves of the split are differently filtered
versions of the sky and hence not expected to match.
Is this jackknife unique in that regards ? I think there are
other jackknives that failed this criterion badly, such as the
ones with partially overlapped scans.
Regardless, this can be fixed by removing the same ground template
from the 2 halves.
11:35 a.m.
New subject: Analysis note: half-scan jackknife and possible culprit
Hi Chao-Lin,
Regardless, this can be fixed by removing the same
ground template
from the 2 halves.
I do remove the same ground template from the two halves, and that's
exactly the reason for concern. This means that information from the
late half-scans goes into the early half-scan map, and vice versa.
Since the ground subtraction procedure mixes the signal between the two
subsets, it will introduce a ground-fixed anticorrelation even if there
was no correlation to start with. This can also be described as
filtering the sky differently in the two subsets, in a way that could
introduce jack failures. In our standard mapmaking we apply ground
subtraction separately to the left-going and right-going scans. There
are other reasons why we do this, but among those reasons it protects
the scan-direction jackknife test from this problem.
The safer thing would be to make separate ground templates for the two
halves. This would potentially make the jackknife maps less directly
comparable with our standard coadded maps, but that's OK. I can
compare them with jack0 coadds and sims using the same hacked code.
- Walt
1:56 p.m.
New subject: Analysis note: half-scan jackknife and possible culprit
Hi Walt and Chao-Lin,
It is our az-fixed scan strategy which makes the filtering due to p3 and
ground subtraction significantly different on the sky for the first half
of the scanset vs. the second half (for which the sky has moved, on
average, by 7 degrees). Sims will show us this effect. This is
different than applying the ground subtraction separately to left- and
right-going scans--these may have different scan-fixed contamination but
the sky enters into them exactly the same, so the filtering of the sky
signal in left vs right is identical.
It's not clear to me that making separate ground templates for the first
half and second half will make this issue of different sky filtering go
away, and it will certainly result in dramatically increased filter
suppression at ell=50. But as you say if you judge your jack0 and alt
jacks against sims it is a fair test.
John
On 2/9/13 5:35 AM, Walt Ogburn wrote:
Hi Chao-Lin,
--
___________________________________________________________________
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
CURRENTLY AT SOUTH POLE STATION, ANTARCTICA
Regardless, this can be fixed by removing the
same ground template
from the 2 halves.
I do remove the same ground template from the two halves, and
that's
exactly the reason for concern. This means that information from the
late half-scans goes into the early half-scan map, and vice versa.
Since the ground subtraction procedure mixes the signal between the two
subsets, it will introduce a ground-fixed anticorrelation even if there
was no correlation to start with. This can also be described as
filtering the sky differently in the two subsets, in a way that could
introduce jack failures. In our standard mapmaking we apply ground
subtraction separately to the left-going and right-going scans. There
are other reasons why we do this, but among those reasons it protects
the scan-direction jackknife test from this problem.
The safer thing would be to make separate ground templates for the two
halves. This would potentially make the jackknife maps less directly
comparable with our standard coadded maps, but that's OK. I can
compare them with jack0 coadds and sims using the same hacked code.
- Walt
_______________________________________________
Bicep2-list mailing list
Bicep2-list(a)lists.fas.harvard.edu
https://lists.fas.harvard.edu/mailman/listinfo/bicep2-list
2:36 p.m.
New subject: Analysis note: half-scan jackknife and possible culprit
Hi John,
It is our az-fixed scan strategy which makes the
filtering due to p3
and ground subtraction significantly different on the sky for the
first half of the scanset vs. the second half (for which the sky has
moved, on average, by 7 degrees).
I agree this is the main point: the ground subtraction template is
constructed in ground-fixed coordinates, and the jackknife is done in
celestial coordinates. This means that the effects of ground
subtraction don't drop out in the jackknife I've constructed.
Sims will show us this effect.
Also agreed, and this is what I have in mind to do next.
This is different than applying the ground subtraction
separately to
left- and right-going scans--these may have different scan-fixed
contamination but the sky enters into them exactly the same, so the
filtering of the sky signal in left vs right is identical.
I think there should be a small difference in filtering of the sky even
for the left/right split. It will, of course be a much smaller
difference than in the first-53 / last-53 jackknife split I am looking
at now.
It's not clear to me that making separate ground
templates for the
first half and second half will make this issue of different sky
filtering go away, and it will certainly result in dramatically
increased filter suppression at ell=50.
I agree it will result in increased filter suppression, which will make
it unsafe to compare my jackknife maps with standard coadds. If my
current construction fails as badly for signal+noise sims as for real
data, then I expect changing to a first-half and last-half ground
subtraction template will make the sims pass. But again, the important
thing is that an apples-to-apples comparison of sims and real data is
the standard of proof.
- Walt
9:08 p.m.
New subject: Analysis note: half-scan jackknife
Hi all,
I have completed enough of a signal-only sim to see that the jackknife
failures in my note are indeed the result of having performed ground
subtraction across the entire scanset. The sim has a signal in the
differenced temperature map that's just as striking as in the real
data. I conclude that the signals in EE, BB, and TB speclin have
similarly been filtered in at low ell from the non-jackknife maps.
Therefore, we can't conclude anything from the jack as I have
constructed it.
http://bicep0.caltech.edu/~spuder/analysis_logbook/analysis/20130207_scanja…
- Walt
9 Feb
9 Feb
7:21 a.m.
New subject: Analysis note: half-scan jackknife
While it is disappointing, we are not back to square one. Walt has identified a feature in
BICEP2 EE that is clearly systematics and not observed in sims or by Keck. One more hint
for the ongoing detective work..
On Feb 8, 2013, at 6:08 PM, Walt Ogburn <ogburn(a)stanford.edu> wrote:
Hi all,
I have completed enough of a signal-only sim to see that the jackknife
failures in my note are indeed the result of having performed ground
subtraction across the entire scanset. The sim has a signal in the
differenced temperature map that's just as striking as in the real
data. I conclude that the signals in EE, BB, and TB speclin have
similarly been filtered in at low ell from the non-jackknife maps.
Therefore, we can't conclude anything from the jack as I have
constructed it.
http://bicep0.caltech.edu/~spuder/analysis_logbook/analysis/20130207_scanja…
- Walt
_______________________________________________
Bicep2-list mailing list
Bicep2-list(a)lists.fas.harvard.edu
https://lists.fas.harvard.edu/mailman/listinfo/bicep2-list
4116
days inactive
4117
days old
bicep2-list@lists.fas.harvard.edu
6 comments
4 participants
participants (4)
-
Chao-Lin Kuo -
Clem Pryke -
John Kovac -
Walt Ogburn