On Fitting Manuela Cirilli Manuela Cirilli
Mar 27, 2015
On Fitting
Manuela CirilliManuela Cirilli
Roma, 25 novembre 2002 Manuela Cirilli FIT/2
Global 6 parameters fit:Global 6 parameters fit: 2 Fermi-Dirac functions with fixed KT constant background
Global 8 parameters fit:Global 8 parameters fit: 2 Fermi-Dirac functions with KT as free parameters constant background
Rising and falling edge fit:Rising and falling edge fit: 1 Fermi-Dirac with KT as free parameters + constant background
Fit options:Fit options: 2 fit ? Log likelihood?
It is crucial to establish a reliable and agreed on It is crucial to establish a reliable and agreed on methodmethod
Global 6 parameters fit:Global 6 parameters fit: 2 Fermi-Dirac functions with fixed KT constant background
Global 8 parameters fit:Global 8 parameters fit: 2 Fermi-Dirac functions with KT as free parameters constant background
Rising and falling edge fit:Rising and falling edge fit: 1 Fermi-Dirac with KT as free parameters + constant background
Fit options:Fit options: 2 fit ? Log likelihood?
It is crucial to establish a reliable and agreed on It is crucial to establish a reliable and agreed on methodmethod
TDC spectrum fits
Roma, 25 novembre 2002 Manuela Cirilli FIT/3
Standard 2 option Standard 2 option
6 parameters fit
Tmax=1305T0= 408.1 Tmax=1305T0= 408.1
2 steps: Bounded fit Log likelihood
2 steps: Bounded fit Log likelihood
Tmax= 1308T0= 407.7 Tmax= 1308T0= 407.7
Roma, 25 novembre 2002 Manuela Cirilli FIT/4
Standard 2 option Standard 2 option
6 parameters fit
Tmax= 1346T0= 416.7 Tmax= 1346T0= 416.7
2 steps: Bounded fit Log likelihood
2 steps: Bounded fit Log likelihood
Tmax= 1350T0= 414.7 Tmax= 1350T0= 414.7
Roma, 25 novembre 2002 Manuela Cirilli FIT/5
6 parameters fit
The problem lies in the KT values The problem lies in the KT values
Roma, 25 novembre 2002 Manuela Cirilli FIT/6
A better way: 8 parameters fit
Better 2/ndf Less dependence on the fit options This is implemented in calib, together with automatic search of input parameters to the fit Works really fine!
…is this the end of the story?
Better 2/ndf Less dependence on the fit options This is implemented in calib, together with automatic search of input parameters to the fit Works really fine!
…is this the end of the story?
Roma, 25 novembre 2002 Manuela Cirilli FIT/7
Let’s try and fit T0 and Tmax separately
Roma, 25 novembre 2002 Manuela Cirilli FIT/8
Difference in T0
T0(8 parameters) - T0(rising edge) Look at a good-behaved chamber (BIS)
T0(8 parameters) - T0(rising edge) Look at a good-behaved chamber (BIS)
ML2ML2ML2ML2
T0=1.9 counts
T0=1.9 counts
ML1ML1ML1ML1
T0=2.06counts
T0=2.06counts
Roma, 25 novembre 2002 Manuela Cirilli FIT/9
Difference in Tmax
Tmax(8 parameters) - Tmax(rising edge) Look at a good-behaved chamber (BIS)
Tmax(8 parameters) - Tmax(rising edge) Look at a good-behaved chamber (BIS)
ML2ML2ML2ML2
Tmax=-2.2 counts
Tmax=-2.2 counts
ML1ML1ML1ML1
Tmax=-2.6 counts
Tmax=-2.6 counts
Roma, 25 novembre 2002 Manuela Cirilli FIT/10
Stability of the T0 fit
Wide region of stabilityWide region of stability for regular spectrafor regular spectra
Wide region of stabilityWide region of stability for regular spectrafor regular spectraT0 Tfit 2/ndf
406.1 430 4.98
405.6 440 4.5
405.5 450 4.4
405.5 460 4.25
405.6 470 4
405.6 480 3.7
405.6 490 3.5
405.6 500 3.48
405.5 510 3.38
Roma, 25 novembre 2002 Manuela Cirilli FIT/11
Stability of the Tmax fit
Wide region of stabilityWide region of stability for regular spectrafor regular spectra
Wide region of stabilityWide region of stability for regular spectrafor regular spectra
Tmax Tfit 2/ndf
1335 1210 31
1342.0 1220 1.15
1342.1 1230 1.04
1342.3 1240 1.02
1342.9 1250 0.91
1343.1 1260 0.91
1343.3 1270 0.87
1343.1 1280 0.89
1343.8 1290 0.88
1343.6 1300 0.95
1344.2 1310 0.96
Roma, 25 novembre 2002 Manuela Cirilli FIT/12
The separate fit of T0 and Tmax seems the best The separate fit of T0 and Tmax seems the best optionoption Stability studies to be done on large number of Stability studies to be done on large number of runs/chambers and on high rate spectraruns/chambers and on high rate spectra Study the cases of not well behaved spectraStudy the cases of not well behaved spectra
The separate fit of T0 and Tmax seems the best The separate fit of T0 and Tmax seems the best optionoption Stability studies to be done on large number of Stability studies to be done on large number of runs/chambers and on high rate spectraruns/chambers and on high rate spectra Study the cases of not well behaved spectraStudy the cases of not well behaved spectra
Conclusions