Standard analysis of MAGIC data with MARS V151

1
Standard analysis of MAGIC data with MARS V1-5-1

• Daniel Mazin and Abelardo Moralejo
• IFAE

2
Timetable of the course
Mazin (analysis)

• Last 2 / 3 hours D. Mazin on spectral unfolding
  skymaps

3
If you are getting involved in MAGIC data
analysis

• Subscribe to the magic_soft mailing list! (info
  on releases, problems in data, bugs)

4
Versions of MARS

• To avoid misunderstandings since March 2005
  there are in MAGIC two versions of the analysis
  code (MARS)
• The official one, whose CVS is maintained by IFAE
  (cvsmagic.pic.es), and which runs at La Palma and
  at the PIC data center
• The one maintained by the Würzburg group
• Most people inside MAGIC use the standard
  version, the one we will present in this course.

5
Analysis handbook

• A complete guide to the standard analysis of
  MAGIC data is under preparation (better late than
  never).
• Please download the preliminary version at

• http//personal.ifae.es/moralejo/web/
• analysisHandbook.pdf
• It will be of help during the course,
  particularly for the real beginners in MARS

6
MARS V1-5-1

• Released 20 / 02 / 2008
• To be run under root 5.12.00g (or f)
• First version prepared to analyze easily the
  wobble data using timing and source-dependent
  parameters
• First version with a standard skymap utility
• Available at CVS, or through the PIC webpage
• http//magic.pic.es/datacenter/cvs/cvs.html

7
MAGIC data center at PIC
8
MAGIC data files

• .raw files contain all the information of a
  triggered event (all FADC slices of all pixels,
  UTC,)
• Three types of raw files pedestals,
  calibrations, and data (showers) file names
  contain the tags _P_, _C_ or _D_
• Data raw files also contain interleaved
  calibration and pedestal events, to track the
  evolution of conversion factors and FADC baseline
• Central control reports (.rep) store
  information from the different subsystems (anode
  currents, drive system, star guider).

9
MAGIC data analysis chain
7 executables one root macro

• callisto ? phe- and arrival time in each pixel
• merpp ? Add information from central control
• star ? Image parameters (shape, orientation)
• osteria ? Optimize ?/h tagging, DISP and E estim.
• melibea ? Calculate hadronness, DISP and E
• fluxlc ? Apply cuts, get spectrum and light curve
• CombUnfold.C ? Unfold spectrum (limited E
  resolution)
• celestina ? Sky maps
• Each one takes as input the output of the
  previous one(s)

10
Auxiliary programs

• mars now reduced to a simple event display
  (reads in callisto _Y_ outputs)
• showplot reads in files containing status
  displays (example calibxxxx.root,
  signalxxxx.root, starxxxx.root) created (besides
  their main output) by the various executables,
  and displays summary plots for the analyzed set
  of files.

11
Input and output files

• callistomerpp _P_.raw , _C_.raw ,
  _D_.raw ? One _Y_.root file per data run
• star _Y_.root ? One _I_.root file per data
  run
• osteria _I_.root (both MC and real data) ?
• RF.root, EN.root, DispPar.root (result of
  optimizations)
• melibea _I_.root ? One _Q_.root file per
  data run
• fluxlc _Q_.root ? spectr_.root (spectrum
  LC)
• CombUnfold.C spectr_.root ? just status display
  (spectrum)
• celestina _Q_.root ? skyplot.root (sky map)
• Status display a graphical output of all the
  executables, which is stored in a root file and
  can be re-read with the program showplot

12
How to use the executables

• Most of the settings of the analysis are written
  inside configuration files callisto.rc, star.rc
  (text files, see Mars/ directory)
• The rest of the options are set directly in the
  command line input and output paths,
• --log A brief set of instructions is obtained
  by running any executable with just the option
  -h

13
Some common command line options (callisto, star,
osteria, melibea, fluxlc)

• -v verbose level (default 5)
• -f force overwrite of output from previous
  executions
• -b batch execution (no graphics)
• -q quit after execution
• -h help
• --logfilename.log direct text output to a
  file
• --configxxxx.rc set configuration file
• (run executables with -h to see all options)

14
Example star.rc file
MJStar.MImgCleanStd.CleanLevel1
10 MJStar.MImgCleanStd.CleanLevel2
5 MJStar.MImgCleanStd.CleanMethod Absolute
MJStar.MHillasCalc.IdxIsland 0 MJStar.MSrcPosCalc
.SourceRaDec 5.5755 22.01444 MJStar.CalibStargu
ider yes MJStar.UseStarguider
yes MJStar.MCalibrateStarguider.Method
Culmination MJStar.MPointingPosInterpolate.Correct
Zenith yes MJStar.MPointingPosInterpolate.Correct
Azimuth yes NOTE Lines starting with are
deactivated (BUT default value of the
corresponding setting may be anything check
documentation of class at cvs.ifae.es in case
of doubt)
15
Sequence files

• Sequences are groups of data files which are
  analyzed together. A sequence is normally defined
  by the availability of a pedestal and a
  calibration file, and includes all _D_ files that
  follow until the next pedestal file.
• Typically, one or few sequences contain the whole
  dataset for a source in a night.
• A sequence file is a short text file containing
  the run numbers of all the files in the sequence.
  This is a needed input for callisto and star. See
  for instance /nfs/magic-calib00/CrabNebula/seqdata
  //seqtxt at PIC

16
Example sequence file

• Sequence 101042
• Night 2006-09-21
• CalRuns 101042
• PedRuns 101041
• DatRuns 101043 101044 101045 101046 101047
  101048 101049 101050 101051 101052
• The last argument of the calls to callisto, star
  and melibea is the name of such a sequence file,
  like sequence00101042.txt

17
Example of star call

• star --ind./ --out./star_tc10_5
• --logstar.log --config./star.rc
  sequence00101042.txt
• First two flags set the input (contains _Y_
  files) and output directories
• --log sets name of text file where the log of
  the execution is written
• --config points to the configuration star.rc file

18

• Nowadays, the first steps of the analysis
  (calibration and image parameterization) are run
  automatically in La Palma and at PIC, with our
  current understanding of the best settings for
  callisto and star
• The analyzer should then start his work by
  downloading the star (_I_.root) files
  corresponding to the data sample he is interested
  in.

19
Where to get star files for the standard analysis

• Available at the PIC data center (usually a few
  days after data taking - please report missing
  data!). Access through magic-ui.pic.es
• Real data /nfs/magic/DataCenter/Analysis/
• Star_time 6-3 phe time cleaning, timing
  parameters
• Star_notime 10-5 phe normal cleaning, no timing
  parameters
• star Monte Carlo files under
• /nfs/magic-montecarlo/MonteCarlo/
• Very important parameters to choose the right MC
• FADC type (3 epochs Siegen, Siegen w/splitters,
  MUX)
• Wobble / no wobble
• Optical PSF of mirror, zenith angle (zbin),
  cleaning

20
Optical PSF evolution
http//magic.pic.es/plot.htm
21
Role of Monte Carlo simulation

• After star, MC files are needed to proceed with
  the analysis
• For training the methods (DISP, energy
  estimation, ?/hadron tagging) train sample
• For estimating the efficiency of the analysis for
  gamma-rays ( collection area) test sample
• Obviously, all the analysis steps (cuts for
  instance) must be the same as for the real data

22
Splitting of MC gamma sample

• Obviously, the MC gamma sample used for the
  training must be independent of the one used for
  estimating collection areas
• Therefore, we must split the MC samples into two
  equivalent ones, usually called train and
  test. This is up to the analyzer (what fraction
  of MC to dedicate to each subsample)

23
Source-dependent and source-independent analysis

• Two approaches in the analysis of MAGIC data
  after image parameterization
• Make use of the a priori knowledge of the
  position of the (point-like) source Dist and
  other source-dependent parameters can be used in
  the g/h separation E-estimation
• No assumption on the source position no
  source-dependent parameter allowed, DISP used to
  estimate the direction of each event ? sky map
• The two approaches are useful, and running them
  in paralel is recommended

24
Running osteria

• Real data star files are used as hadrons for
  training the ?/h tagging, together with MC
  gammas. DISP and E estimation training require
  just MC gammas (the used MC files are set
  independently for each task).
• The different optimizations can be run
  separatedly
• osteria --rf --rfhadronin2007_I_root
  --rfgammafile Gamma_I_root
• osteria --disp --dispgammafileGamma_I_roo
  t
• osteria --train-enr --rfengammafileGamma_
  I_root
• Check and try the runosteriacsh scripts provided
  for the course together with the example data set
• magic-ui.pic.es/nfs/magic-buffer00/BenasqueSchoo
  l

25
Cuts prior to training

• Some basic cuts can be set in osteria.rc
• SkipNonShowers sparks, minimum SIZE
• FilterCuts leakage, of core pixels, number of
  islands
• Note that cut expressions in osteria.rc define
  the events to be rejected
• Beware for wobble data, do not make cuts in
  parameters which are source-dependent (contained
  in MHillasSrc)
• Cuts will be automatically passed down the
  analysis chain. Filter cuts can be modified in
  melibea (only for test purposes!)

26
Example runosteria.csh

• MARSSYS/osteria -f -q \
• --configMARSSYS/osteria.rc \
• --rfhadronin./OFF/200root" \
• --rf --rfgammafile./MC/train/Gammaroot" \
• --rfout. --ntrees100 --zdmax30. \
• --log./osteria.log

Trains Random Forest for gamma/hadron separation
(--rf) growing 100 trees. With the default
osteria.rc file, both the source-dependent and
source-independent approach are trained.
27
Setting ?/h separation and energy estimation
parameters in osteria.rc. Source-dependent

• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn0 log10(MHillas.fSize)
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn1 2.55floor(MPointingPos.fZd/5)
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn2 MHillas.fWidth
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn3 MHillas.fLength
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn4 log10(MHillas.fSize/(MHillas.fWidthMHilla
  s.fLength))
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn5 MNewImagePar.fConc
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn6 MHillasSrc.fDist
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn7 MHillasExt.fM3Longsgn(MHillasSrc.fCosDelt
  aAlpha)
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn8 MHillasTime.fRMSTime
• OsteriaLoop.MRFGHSeparation.RFGHSeparationMatrix.C
  olumn9 MHillasTimeFit.fP1Grad
• sgn(MHillasSrc.fCosDeltaAlpha)
• OsteriaLoop.MRFEnergyEst.RFEnergyEstMatrix.Column0
  log10(MHillas.fSize)
• OsteriaLoop.MRFEnergyEst.RFEnergyEstMatrix.Column1
  MHillasSrc.fDist
• OsteriaLoop.MRFEnergyEst.RFEnergyEstMatrix.Column2
  MHillas.fWidth
• OsteriaLoop.MRFEnergyEst.RFEnergyEstMatrix.Column3
  MHillas.fLength
• . (5 more parameters)
• OsteriaLoop.MRFEnergyEst.RFEnergyEstMatrix.Column9
  MMcEvt.fEnergy
• Column numbering must start from 0 and numbers
  must be consecutive. For energy training, last
  column is the true MC energy

28
Setting ?/h separation and energy estimation
parameters in osteria.rc. Source-independent

• OsteriaLoop.MRFGHSeparationSrcIndep.RFGHSeparation
  Matrix.Column0 log10(MHillas.fSize)
• OsteriaLoop.MRFGHSeparationSrcIndep.RFGHSeparation
  Matrix.Column1 2.55floor(MPointingPos.fZd/5)
• OsteriaLoop.MRFGHSeparationSrcIndep.RFGHSeparation
  Matrix.Column2 log10(MHillas.fSize/(MHillas.fWidt
  hMHillas.fLength))
• OsteriaLoop.MRFGHSeparationSrcIndep.RFGHSeparation
  Matrix.Column3 MHillas.fWidth
• OsteriaLoop.MRFGHSeparationSrcIndep.RFGHSeparation
  Matrix.Column4 MHillas.fLength
• OsteriaLoop.MRFGHSeparationSrcIndep.RFGHSeparation
  Matrix.Column5 MNewImagePar.fConc
• OsteriaLoop.MRFGHSeparationSrcIndep.RFGHSeparation
  Matrix.Column6 MHillasTime.fRMSTime
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column0 log10(MHillas.fSize)
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column1 MHillas.fWidth
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column2 MHillas.fLength
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column3
• log10(MHillas.fSize/(MHillas.fLengthMHillas.fWidt
  h))
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column4 MNewImagePar.fConc
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column5 MNewImagePar.fLeakage1
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column6 MPointingPos.fZd
• OsteriaLoop.MRFEnergyEstSrcIndep.RFEnergyEstMatrix
  .Column7 MMcEvt.fEnergy

29
Output of osteria

• Files containing Random forests
• RF.root
• RFSrcIndep.root
• Besides the random forests, these files contain
  the status displays of the osteria execution can
  be opened using showplot

30
showplot RF.root
31
c.o.g. (wobble data)

• MC gamma mostly on a ring around the simulated
  source location
• Hadrons (real data) inhomogeneities

32
Training hadron sample

• Note usually we can take as training hadron
  sample the same data (ON or wobble) we want to
  analyze. When dealing with a strong source (like
  Crab), due to the gamma content of the data, some
  improvement in performance may result from using
  other data for the training, containing no gamma
  source.
• Of the hadron events provided to osteria
  through --rfhadronin , only a small subsample
  (amounting to as many events as MC gamma events)
  is used. It is selected automatically (and by
  default) when running osteria, such that its
  overall SIZE and zenith angle distributions
  matches the ones of gammas.

33
z.a. versus log10(Size) before training hadron
selection (in the Events vs. z.a. vs. Size tab)
34
z.a. versus log10(Size) after training hadron
selection
Aim avoid that RF uses SIZE or z.a. directly as
a ?/h discriminators (they are useful as scaling
parameters, on which others depend, but in
themselves are not a good discriminators)
35
Importance of RF parameters

• Note relevance of parameter depends on Size
  range for Sizegt80 phe- (this case) Dist is
  strongest. At higher Sizes W, L,
• This plot very useful to detect possible
  mistakees in the selection of the MC (if a
  parameter is too good!)

36
Fast check of g/h separation (source-dep)

• OsteriaLoop.MRFGHSeparation.TrainRatio 0.95
• OsteriaLoop.MRFEnergyEst.TrainRatio 0.95 (in
  osteria.rc)

37
Hadronness vs. Size (source-dep)
hadrons
?
38
Source-independent ?/h separation

• showplot RFSrcIndep.root

Note head-tail assymetry separation inside DISP
method!
39
Example runosteria_EandDisp.csh

• MARSSYS/osteria -f -q \
• --configMARSSYS/osteria.rc --zdmax30. \
• --train-enr --rfengammafile./MC/train/Gammaroot
  " \
• --rfout. --ntrees50 \
• --disp --dispgammafile./MC/train/Gammaroot" \
• --dispout./ \
• --log./osteria_EandDisp.log

Trains Random Forest for energy estimation
(--train-enr) growing 50 trees. With the default
osteria.rc file, both the source-dependent and
source-independent approaches are trained. DISP
method is also optimized (--disp)
40
Output of osteria (E and DISP optimization)

• Files containing energy Random forests
• EN.root
• ENSrcIndep.root
• File containing DISP parametrization
• DispPar.root
• Each of them contains a (rather incomplete)
  status display

41
DispPar.root

• gamma-PSF transversal and longitudinal (along
  image axis)

42
DispPar.root (leakage)

• gamma-PSF transversal and longitudinal (along
  image axis)

43
EN.root, ENSrcIndep.root

• Only plot Etrue vs. Eest
• Source-dep approach clearly better

44
melibea

• melibea just applies to the _I_ files the same
  cuts used in osteria for the training, and then
  calculates the estimated energy and the
  hadronness (and DISP for the source-independent
  analysis)
• Source-dependent approach using the RF.root and
  EN.root outputs of osteria).
• Source-independent approach using
  RFSrcIndep.root, ENSrcIndep.root, DispPar.root
• The melibea output is one _Q_ file per _I_
  file, plus a melibea.root status display

45
melibea.rc

• For standard analysis you can use an empty file
  as melibea.rc (check the one in
  BenasqueSchool/rcFiles). In this way, the cuts
  used for training will be automatically applied
  by melibea (Size, sparks, ).

46
Example runmelibea.csh(source-dep analysis)

• MARSSYS/melibea -q -f \
• --config./rcFiles/melibea.rc \
• --ind./Crab --out./Crab/melibea \
• --srcCrabNebula --rf --rftree./RF.root \
• --calc-enr --rfentree./EN.root \
• --UseDefaultSourceRADec --NOffWobble3 \
• --log./Crab/melibea/melibea.log

47
Example runmelibea.csh

• --srcCrabNebula process CrabNebularoot star
  files
• --rf --rftree./RF.root calculate hadroness
  using the random forest in RF.root
• --calc-enr --rfentree./EN.root calculate
  estimated energy using the random forest in
  EN.root
• For wobble mode only
• --UseDefaultSourceRADec --NOffWobble3
  re-calculate MHillasSrc, hadronness and energy
  (all source-dep) with respect to 3 Off
  positions on the camera

48
Besides, with --NOffWobble3

• The melibea (_Q_.root) files have the same
  contents of their progenitor star files (for the
  events surviving cuts), plus
• Mhadronness.fHadronness
• MEnergyEst.fEnergy

• MHadronness180.fHadronness, MEnergyEst180.fEnergy,
  
• MHillasSrc180.
• MHadronness090.fHadronness, MEnergyEst090.fEnergy,
  
• MHillasSrc090.
• MHadronness270.fHadronness, MEnergyEst270.fEnergy,
  
• MHillasSrc270.

• Note option --NOffWobble re-defined since Mars
  V1-5-0!

49
Source-dependent wobble data analysis (Mars gt
V1-5-0)
090
MHillasSrc090.fAlpha MEnergyEst090.fEnergy MHadron
ness090.fHadronness
MHillasSrc.fAlpha MEnergyEst.fEnergy MHadronness.f
Hadronness
source
Camera center
180
MHillasSrc180.fAlpha MEnergyEst180.fEnergy MHadron
ness180.fHadronness
MHillasSrc270.fAlpha MEnergyEst270.fEnergy MHadron
ness270.fHadronness
270
50
Running melibea on MC example runmelibea_MC.csh

• MARSSYS/melibea -mc -q -f --config./rcFiles/meli
  bea.rc \
• --ind./MC/test \
• --out./MC/test/melibea/ \
• --srcGamma --rf --rftree./RF.root \
• --calc-enr --rfentree./EN.root \
• --NOffWobble3 \
• --log./MC/test/melibea/melibea.log
• One must use the same RF.root, and EN.root as for
  the real data. The MC files to be melibeated are
  the test files, that is, they must be
  independent of those used for the trainings in
  melibea.

51
Some exercises on the melibea files (mainly for
beginners)

• Using roots TChain on the Events tree (see some
  hints on how to use it in the analysis handbook
  draft)
• Make some Alpha plots with tentative Size /
  hadronness cuts, for the on and (overlaid) for
  the wobble offs
• Do the same for theta2 plots (using the _Q_ files
  with DISP, and the coordinates of the source
  position on the camera)
• With the MC melibea files, check what fraction of
  the signal may end up in the off at low Alpha,
  after the hadronness cut (for the 3 off positions)

52
melibea.root file

• This is an additional output of melibea
  containing a status display
• Main contents
• Histograms of various parameters (in MHillas,
  MHillasExt, MHillasSrc)
• Plots of efficiency of prior cuts (Size, zenith
  angle, sparks) vs. Time (both overall and in
  Size bins) this may be helpful to find bad runs
  (for this purpose, check also the starroot files
  which contain the status display of star).

53
Energy spectrum and light curve

• At this point we must have a set of real data
  melibea files , i.e. 2007_Q_.root, and a set of
  gamma MC melibea files, i.e. Gamma_Q_.root
• Now we have to run fluxlc. Everything in fluxlc
  is configured via fluxlc.rc, then command line is
  just
• fluxlc --configfluxlc.rc --logflux.log
• Check MARSSYS/fluxlc.rc

54
fluxlc.rc

• Main things to be defined
• paths to files, output filename
• Number of z.a. bins (for now, choose 1) and z.a.
  range.
• Number of bins in energy (and E range)
• Wobble / no wobble
• Alpha (source-dep) or Theta2 (source-indep)
  analysis
• Set hadronness and Theta2 (or alpha) cuts
• Timescale of lightcurve (hours, days)

55
fluxlc.rc

• Setting the files to be used, and the output file
  name
• FluxLC.mcdata ../MC/test/melibea/Gamma_Q_.root
  
• FluxLC.data ../Crab/melibea/200_Q_.root
• FluxLC.offdata (only for on/off observations)
• FluxLC.outname spectr_calc_out_Crab.root
• Zenith angle range
• FluxLC.zabins 1 (do not change)
• FluxLC.za_edges 0., 29.

56
fluxlc.rc

• Setting additional cuts (beyond those applied by
  osteria/melibea) if needed
• FluxLC.UserCuts (MNewImagePar.fLeakage1lt0.2)
  (MImagePar.fNumIslandslt2)
• Note no source-dep parameters in wobble data! A
  harder leakage cut, or various other cuts may be
  useful cut to test the stability of features in
  the spectrum or light curve.
• FluxLC.MaxDist 1.2(x-2.5)0.2
• Will be applied for each off position in wobble
  data, in order not to break symmetry. x stands
  for log10(Size). We can also set a fixed maximum
  Dist.
• FluxLC.MinSize 60 (phe)

57
fluxlc.rc

• Setting some binnings
• FluxLC.nBinsEnergyEst 30
• FluxLC.lowerEest 5.
• FluxLC.upperEest 50000.
• Data type
• FluxLC.WobbleData TRUE
• Type of analysis
• FluxLC.AngleType Alpha (source-dep analysis)
• FluxLC.AngleType Theta2 (source-indep analysis)

58
How to choose the hadronness and Alpha (or ?2)
cuts?

• A compromise is needed between statistical errors
  and systematic errors
• Too loose cuts bad BG discrimination, poor
  significance ? large error bars in spectrum
• Too tight cuts statistical significance may be
  good, but at the expense of introducing a larger
  systematic error in the collection area due to
  the possible differences between the MC gammas
  and the real ones. Tight cuts will keep only
  those real events which are most similar to the
  MC

59
Two ways of setting hadronness and Alpha cuts

• FluxLC.FindCutsFromEfficiency TRUE
• FluxLC.AlphaEffi 0.8
• FluxLC.HadEffi 0.5
• The program will try to set cuts such that the
  requested efficiency is reached in each bin of
  Size. There are minimum allowed values in
  current Mars minimum values of the cuts are set
  by default at 8 degree and 0.01
• FluxLC.FindCutsFromEfficiency FALSE
• FluxLC.had 0.2, 0.2, 0.2, (hadronness cut)
• FluxLC.alp 8.,8.,8., (Alpha or theta2 cut)
• ? Cuts set manually one by one in each of the
  bins of estimated energy

60
Wobble data how many off regions?

• FluxLC.NumberOfWobbleOff 1
• Using just the antisource (1) is the safest
  choice to ensure an unbiased background
• FluxLC.NumberOfWobbleOff 3
• The statistical error in the determination of
  background will be smaller (3? statistics) , but
  the two additional Off positions are not totally
  equivalent to the source position (possible
  systematics due to camera inhomogeneities!).
  Still, it should be safe above 200 GeV

61
Background normalization

• Wobble use just the geometrical factor
  (possible gamma contamination in the OFF)
• FluxLC.WobGeomNorm TRUE
• FluxLC.WobLCGeomNorm TRUE
• (1/1) or (1/3)
• No wobble normalization by comparing on and off
  out of the signal region. Two alternatives
• FluxLC.CommonNormFactor TRUE
• all energy bins have the same norm. factor
• FluxLC.CommonNormFactor FALSE
• each energy bin is normalized independently

62
Light curve settings

• FluxLC.Lightcurve TRUE
• FluxLC.LowELC 300. (GeV)
• FluxLC.UpELC 50000. (GeV)
• FluxLC.AutotimeBinningLC TRUE
• FluxLC.LCTimeScale 10.
• (min) The program wiill try to do bins in time of
  such duration
• FluxLC.AlphaCutForLC 8.
• FluxLC.Theta2CutForLC 0.04
• FluxLC.TimeFormat HMS (or dmy, MJD)

63
Calculation of collection areas

• Obtained from MC in fine bins of zenith angle and
  energy, then weighted according to the z.a.
  distribution of the data, and to an assumed
  energy distribution (default -2.6 spectral index)
  in order to obtain the coll. Area in the coarse
  bins (E and z.a.) of the analysis
• For the spectrum unfolding done later by the
  program following fluxlc
• For the light curve since V1-5-0, a sort of
  spillover correction is applied to account for
  the finite energy resolution.

64
Recommendations for running fluxlc

• Managing to obtain a spectrum does not mean it is
  the correct one
• Always try to apply your analysis (RF, energy
  est) to the Crab, from data as similar as
  possible (PSF, z.a., FADC type) to those you are
  interested in. Check the obtained Crab spectrum
  before the one of your source.
• Try different cut strengths (Alpha, hadronness).
  In this way you can estimate the possible
  systematics due mismatches of MC and data ? grey
  band in spectrum

65
Recommendations for running fluxlc (II)

• Managing to obtain a spectrum does not mean it is
  the correct one
• In wobble, check whether or not the spectrum
  changes significantly by using just 1 off
  (antisource) w.r.t. using, for instance, 3.
• In any case, check always the individual Alpha or
  ?2 plots to look for background estimation
  problems

66
Output of fluxlc
User-selected maximum Dist

• spectr_calc_.root ? input for unfolding
• Status_spectr_calc_-root ? status display

67
Cut efficiency vs. Eest

• Note do not expect exactly the requested
  efficiencies in the case FindCutsFromEfficiency
  TRUE. Those are in Size bins, and there are
  minimum allowed values of the cuts

68
Rate vs. Zenith angle
69
Collection area in fine bins
70
Coll. Area in coarse bins (for spectrum)
71
Overall Alpha distribution
72
Differential energy spectrum

• Not yet unfolded
• Automatic upper limits calculation not yet
  introduced (? introduce Mrolke here?)

73
Coll. Area for light curve
74
Light curve

• BG also shown to identify possible trends

75
Running fluxlc

• Edit the fluxlc.rc file (from MARSSYS for
  instance)
• Set the correct input files for your sample
• (optional) set a MinSize cut of 150phel to speed
  up execution
• Prepare several copies of that fluxlc.rc, with
  different values of cut strengths, with/without
  maximum Dist cut, with 1 or 3 wobble Offs Take
  care to name the output files differently
• Run these to use their outputs for the unfolding
  tomorrow.