Bug #1496

Updated by Knödlseder Jürgen over 9 years ago

When fitting all parameters, the fit errors of the log parabola model are too large.

The problem can be reproduced using this XML file: attachment:crab_logparabola.xml.

The ctlike run produces the following output:
<pre>
2015-07-01T06:03:36: +=================================+
2015-07-01T06:03:36: | Maximum likelihood optimisation |
2015-07-01T06:03:36: +=================================+
2015-07-01T06:03:36: >Iteration 0: -logL=133492.928, Lambda=1.0e-03
2015-07-01T06:03:36: >Iteration 1: -logL=133489.599, Lambda=1.0e-03, delta=3.330, max(|grad|)=3.544902 [Index:8]
2015-07-01T06:03:36: >Iteration 2: -logL=133489.598, Lambda=1.0e-04, delta=0.000, max(|grad|)=0.013988 [Index:8]
2015-07-01T06:03:37:
2015-07-01T06:03:37: +=========================================+
2015-07-01T06:03:37: | Maximum likelihood optimization results |
2015-07-01T06:03:37: +=========================================+
2015-07-01T06:03:37: === GOptimizerLM ===
2015-07-01T06:03:37: Optimized function value ..: 133489.598
2015-07-01T06:03:37: Absolute precision ........: 0.005
2015-07-01T06:03:37: Acceptable value decrease .: 2
2015-07-01T06:03:37: Optimization status .......: converged
2015-07-01T06:03:37: Number of parameters ......: 11
2015-07-01T06:03:37: Number of free parameters .: 6
2015-07-01T06:03:37: Number of iterations ......: 2
2015-07-01T06:03:37: Lambda ....................: 1e-05
2015-07-01T06:03:37: Maximum log likelihood ....: -133489.598
2015-07-01T06:03:37: Observed events (Nobs) ...: 57339.000
2015-07-01T06:03:37: Predicted events (Npred) ..: 57337.998 (Nobs - Npred = 1.00155)
2015-07-01T06:03:37: === GModels ===
2015-07-01T06:03:37: Number of models ..........: 2
2015-07-01T06:03:37: Number of parameters ......: 11
2015-07-01T06:03:37: === GModelSky ===
2015-07-01T06:03:37: Name ......................: Crab
2015-07-01T06:03:37: Instruments ...............: all
2015-07-01T06:03:37: Instrument scale factors ..: unity
2015-07-01T06:03:37: Observation identifiers ...: all
2015-07-01T06:03:37: Model type ................: PointSource
2015-07-01T06:03:37: Model components ..........: "SkyDirFunction" * "LogParabola" * "Constant"
2015-07-01T06:03:37: Number of parameters ......: 7
2015-07-01T06:03:37: Number of spatial par's ...: 2
2015-07-01T06:03:37: RA .......................: 83.6331 [-360,360] deg (fixed,scale=1)
2015-07-01T06:03:37: DEC ......................: 22.0145 [-90,90] deg (fixed,scale=1)
2015-07-01T06:03:37: Number of spectral par's ..: 4
2015-07-01T06:03:37: Prefactor ................: 5.86089e-16 +/- 4.39833e-11 [1e-23,1e-13] ph/cm2/s/MeV (free,scale=1e-16,gradient)
2015-07-01T06:03:37: Index ....................: -2.32353 +/- 4567.39 [-0,-5] (free,scale=-1,gradient)
2015-07-01T06:03:37: Curvature ................: -0.0707068 +/- 0.00274528 [-0.01,-1000] (free,scale=-1,gradient)
2015-07-01T06:03:37: PivotEnergy ..............: 998705 +/- 3.22562e+10 [10000,1e+09] MeV (free,scale=1e+06,gradient)
2015-07-01T06:03:37: Number of temporal par's ..: 1
2015-07-01T06:03:37: Normalization ............: 1 (relative value) (fixed,scale=1,gradient)
2015-07-01T06:03:37: === GCTAModelIrfBackground ===
2015-07-01T06:03:37: Name ......................: Background model
2015-07-01T06:03:37: Instruments ...............: CTA
2015-07-01T06:03:37: Instrument scale factors ..: unity
2015-07-01T06:03:37: Observation identifiers ...: all
2015-07-01T06:03:37: Model type ................: "PowerLaw" * "Constant"
2015-07-01T06:03:37: Number of parameters ......: 4
2015-07-01T06:03:37: Number of spectral par's ..: 3
2015-07-01T06:03:37: Prefactor ................: 0.999265 +/- 0.0215484 [0.001,1000] ph/cm2/s/MeV (free,scale=1,gradient)
2015-07-01T06:03:37: Index ....................: -0.0117382 +/- 0.0123976 [-5,5] (free,scale=1,gradient)
2015-07-01T06:03:37: PivotEnergy ..............: 1e+06 [10000,1e+09] MeV (fixed,scale=1e+06,gradient)
2015-07-01T06:03:37: Number of temporal par's ..: 1
2015-07-01T06:03:37: Normalization ............: 1 (relative value) (fixed,scale=1,gradient)
</pre>

Back