Transcript Update on the PS analysis - Indico

Update on the PS analysis
Claudio Bogazzi
AWG videoconference
15/07/2011
Outline
1)Transition to SeaTray
2)Nhits – PSF correlations
3)Fixed search, Full sky search
Documentation available
• First talk at the collaboration meeting in Amsterdam (Nov. 2010)
• Since then, updates @CERN and @MOSCOW + AWG videoconferences.
• NEW: http://www.nikhef.nl/~claudiob/pnt2/index.shtml
(all the documentation will end up here)
• Internal note available: ANTARES-PHYS-2011-005
• Version 1 submitted on the CDS system on Wednesday (night), 13th.
• Most of the work is included, however still
a couple of things are missing:
1) Full sky search analysis (see later)
2) Systematic uncertainty on Nhits (see also later)
• All the plots / results in the note are made with our aafit
production.
2
Transition to SeaTray
• Adopted our tools to read the mudst files of the SeaTray production.
Top left: run number
distribution for selected events
after a cut Λ > -5.8 and β < 1°
Bottom left: same as before,
but with a smaller range.
243 runs (QB >= 1) not
reconstructed with the
SeaTray production (18
days of livetime in total).
Already informed Thomas,
waiting the recovery.
3
Transition to SeaTray
• All the SeaTray production files already converted to a format “more
familiar” to us.
Top right: cumulative lambda
distribution for selected events
with Λ > -5.8 and β < 1°
Bottom right: same as before,
but with Λ > -5.2
In both plots we consider the
same range of runs
With a more stricter cut
the aafit production has
more events (2864 vs
2804).
Remember: selection of
basic runs different for the
2 productions.
Livetime: 798 days vs 771
4
Transition to SeaTray
•



•
Some numbers for the SeaTray production (mudst file):
Basic runs, NO SCAN: 6903 runs, 771.9 days
Basic runs, NO SCAN, missing: 243 runs, 18.9 days (!)
Basic runs, NO SCAN, no alignment but in SeaTray prod: 46 runs, 5.4 days
These last 46 runs are not available in our aafit production. The reason is
that there is no alignment during the run interval
What happens is the algorithm
looks always for the closest
alignment (for some events will
be the one before the start of
the run, for others the one after
the end of the run)
No reasons to keep these runs
-> cut to reject them
Run duration
Alignment time
Alignment time
5
SeaTray - Rbr MC (zenith)
Λ > -5.0
Λ > -5.2
149 basic runs (13.8 days of livetime) missing in the rbr MC production
(these and the next plots not scaled yet)
6
SeaTray - Rbr MC (lambda)
Fig. 25 in
the note
SeaTray prod. for data
Rbr MC (up nu/anu, down
nu/anu, MUPAGE)
Aafit standalone prod. for data
Old MC (up nu/anu, CORSIKA)
7
SeaTray - Rbr MC (lambda)
Fig. 25 in
the note
Muons (previous slide): mupage overestimates a bit the data in the
high lambda region (not a surprise). Better modeling (ratio < 1.5) .
Neutrinos (this slide): good agreement as before.
8
SeaTray - Rbr MC (lambda)
9
Nhits s
• In order to take into account correlations between the number of hits and
PSF we decide to change a bit the algorithm at the PE generation level.
• Reminder: so far for the PE generation we used two distributions of the
number of hits. One for the background ( data ) and one for the signal (E-2
neutrinos MC)
Black dots - > PE background .
Blue line -> likelihood (background)
Red line -> PE (signal) and likelihood
(signal)
Using the data distribution for the PE bkg
generation is a conservative approach.
All the results in the note are made with this procedure
10
Nhits studies
• Reminder: in the PE generation, a declination dependent distribution of
log(β) is used.
• In the search method instead we
require a smooth parameterization
of the PSF (help the fit to work well)
11
Nhits studies
• Now, for the PE generation, the PSF as well as the number of hits are
extracted from a 3D histogram log(β) vs sin(δ) vs Nhits
Projection of the z-x axis
From the 2D histo we select two random numbers, one for Nhits the other for the PSF
12
Nhits study
• Small change also at the likelihood computation level (for the background)
Λ > -5.2
-- data
-- nu MC
-- mu MC
-- total MC
Included now
muon contribution
on the Nhits
distribution (14%)
Let’s have a look at the first sensitivity plots with these improvements
13
Fixed search
• Btw, found a bug on the way we scaled the MCs with the livetime: 13%
more events NOW
• In the note all the MC vs data plots are not affected by this bug (was
already fixed), but the ACCEPTANCE and of course the SENSITIVITY are.
Λ > -5.2
Value for 2 declinations
(30° and 40°) not yet
included
14
Fixed search
Λ > -5.4 not yet ready (not a big
deal since this cut is not our
first choice)
To be compared with figure 39
in the note.
15
Full sky search
Λ > -5.2 and δ = -70°
Distribution of the test statistic.
Q3σ = 15.90
Q5σ = 25.59
Probability of 3σ (5σ)
discovery as function of the
mean number of signal
events.
16
Conclusions & next steps
• SeaTray transition already started. As soon as all the data are
reconstructed we can definitely use it.
• Rbr MC: first data / MC plots available. Promising results. Need to make a
decision: do we use it or we continue with the old MC?
• Search algorithm adapted to take into account Nhits – PSF correlations.
• Fixed a bug related on the livetime scaling for the MC->13% more evts
• New results for the fixed and the full sky search available (still a couple of
declinations are missing as well as the > -5.4 cut)
• Update the note with all these new results and inform the collaboration.
• Already received comments from Juergen and Juande (already answered
to some questions).
17