Commit 87403836b9c392741a2d33c75f7ac056814c3566
1 parent
0526c208
corrected for use of model in dealing with source extension
Showing
1 changed file
with
23 additions
and
5 deletions
doc/source/users/tutorials/quickstart/onoff.rst
| ... | ... | @@ -13,10 +13,6 @@ Doing an On/Off analysis |
| 13 | 13 | (background) regions. An On/Off analysis is recommended if you |
| 14 | 14 | want to assure minimal dependency on the Monte Carlo background model. |
| 15 | 15 | |
| 16 | - .. warning:: | |
| 17 | - For the time being the **On/Off analysis only works for point sources**. Make | |
| 18 | - sure that the spatial component of the model you want to analyse is of | |
| 19 | - type ``PointSource``. | |
| 20 | 16 | |
| 21 | 17 | Finally, we consider the classical technique for IACT spectral analysis, |
| 22 | 18 | in which 1D spectra for On and Off regions are used jointly to |
| ... | ... | @@ -64,6 +60,21 @@ To derive On/Off observations from the ``events_edisp.fits`` event list, type: |
| 64 | 60 | Declination of source region centre (deg) (-90-90) [22.01] |
| 65 | 61 | Radius of source region circle (deg) (0-180) [0.2] |
| 66 | 62 | |
| 63 | +We have provided in input a model definition file. This serves | |
| 64 | +a twofold purpose: 1) a source model is used to compute the instrument | |
| 65 | +response in the On region taking into account the source morphology | |
| 66 | +and instrument PSF; 2) a background model is used to calculate the | |
| 67 | +background response, namely the expected background rate in each | |
| 68 | +energy bin and the ratio of background rate in the On region over the | |
| 69 | +rate in the Off regions. You can skip using a background model by | |
| 70 | +setting the hidden parameter ``use_model_bkg=no``: in this case the | |
| 71 | +background rate as a function of energy will be determined later | |
| 72 | +during the fit, and the ratio of background rate in the On region over the | |
| 73 | +rate in the Off regions is set to the ratio of the respective solid | |
| 74 | +angles. Furthermore, you can skip passing a model definition file | |
| 75 | +(pass ``NONE``): in this case the response will be calculated for a | |
| 76 | +pointlike source at the centre of the On region. | |
| 77 | + | |
| 67 | 78 | .. note:: |
| 68 | 79 | We have used the events simulated accounting for energy dispersion, since |
| 69 | 80 | energy dispersion is always used in On/Off analysis. |
| ... | ... | @@ -253,7 +264,14 @@ Alternatively, you can use ``WSTAT`` for an On/Off analysis, which treats the |
| 253 | 264 | number of background counts in each energy bin as a nuisance parameter that is |
| 254 | 265 | derived from the On and Off counts by profiling the likelihood function. In |
| 255 | 266 | this case, the only assumption is that the background rate spectrum is the same |
| 256 | -in the On and Off regions. Below the results for a :ref:`ctlike` run with | |
| 267 | +in the On and Off regions. | |
| 268 | + | |
| 269 | +.. note:: | |
| 270 | + You must use ``WSTAT`` if you have selected | |
| 271 | + ``use_model_bkg=no`` in `csphagen` . `csphagen` sets automatically ``WSTAT`` as | |
| 272 | + statistic in the :ref:`observation definition file <glossary_obsdef>` in this case. | |
| 273 | + | |
| 274 | +Below the results for a :ref:`ctlike` run with | |
| 257 | 275 | the ``statistic=wstat`` option. |
| 258 | 276 | |
| 259 | 277 | .. code-block:: none | ... | ... |