Influence of sky pixelisation on False Discovery Rate method for
Transcription
Influence of sky pixelisation on False Discovery Rate method for
Influence of sky pixelisation on False Discovery Rate method for Point Sources Analysis Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles What about ... Sky pixelisation definition Analysis principles FDR Method Influence of the binning on the FDR Conclusions and Outlook Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Sky Pixelisation Binning characteristics, taking into account the symetries of the detector : Constant solid angle : Equal Area Isolatitude pixelisation 3 differents pixelisations : RECTANGLE RECTANGLESINUS RHOMBUS Belgian IceCube Meeting Brussels July 6th 2006 Defined by several parameters Mathieu Labare Université Libre Bruxelles Pixelisation Parametrisation RHOMBUS N ≠ 0 N = 0 Rectangular 1 SINUS 0 RECTANGLESINUS RECTANGLE NROWS SOLID_ANGLE Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Analysis Principles Systematical analysis on 10 000 skies randomized on right ascension Wich binning would be the most efficient for Point Source Analysis ? Parameters optimisation in order to be as close as possible to the required FDR. Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Analysis Principles Compute the amount of neutrinos in a sky bin number ev. backgroung significance nnu[j] Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Analysis Principles Background calculation row(j) number ev. backgroung significance nbg[i] = Avg(nnu[k]) k ∈ row(i) k≠i Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Analysis Principles Compute the significance following a Poisson Law number ev. backgroung significance sign[i] = log(p[i]) ∞ −nbg[i] nbg[i] p[i]= ∑ e x! x=nnu[i] Belgian IceCube Meeting Brussels July 6th 2006 x Mathieu Labare Université Libre Bruxelles Rejection of Null Hypothesis Previously : 3 rejection 54.9 % of rejection Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Rejection of Null Hypothesis Now : FDR Method 1.13 % of rejection Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles FDR vs Solid Angle RECTANGLE Work in progress Restrictions on the parameters ? Belgian IceCube Meeting Brussels July 6th 2006 RHOMBUS N=3 Mathieu Labare Université Libre Bruxelles FDR vs Solid Angle RECTANGLE SINUS Work in progress linear behaviour ? RHOMBUS N=3 Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles FDR vs Solid Angle RECTANGLE SINUS Work in progress linear behaviour ? stability ? error estimation ? ... Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Pseudosources distributions RECTANGLE SINUS 6 4 5 3 5 12 10 3 6 5 11 7 5 18 4 7 Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Pseudosources distributions RECTANGLE 0 4 10 1 4 7 6 9 12 6 8 5 11 7 7 8 Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Pseudosources distributions RHOMBUS 3 9 10 4 7 3 8 7 10 8 9 8 7 9 9 6 8 Belgian IceCube Meeting Brussels July 6th 2006 Mathieu Labare Université Libre Bruxelles Conclusions & Outlook Seems to have some interesting behaviors that have to be well understood : Strong influence of the pixelisation and its parameters Links with the symetries of the detector (solid angle, declination dependance,...) Systematical analysis will continue with others parameters : Shift on the binning Another calculation of the background Incorporation of a simulated Point Source Belgian IceCube Meeting th Brussels July 6 2006 Mathieu Labare Université Libre Bruxelles