Add note about the effect of fitting pivot energies.

Jürgen Knödlseder
1 parent e170765b
Showing 1 changed file with 16 additions and 0 deletions
doc/source/issues.rst
@@ -6,6 +6,7 @@ Known issues
 Below you will find a list of known issues.
 - :ref:`Broken power law has unreliable errors <issues_bplaw>`
+- :ref:`Errors become unreliable when fitting the pivot energy <issues_pivot>`
 .. _issues_bplaw:
@@ -17,3 +18,18 @@ Below you will find a list of known issues.
    and this is related to the fact that the law's gradient is discontinous
    in energy. There is not very much we can do about it, it's inherent in
    the law.
+
+.. _issues_pivot:
+
+.. topic:: Errors become unreliable when fitting the pivot energy
+
+   The spectral ``PowerLaw``, ``ExpCutoff`` and ``LogParabola`` models
+   have a pivot energy, specified by the ``Scale`` parameter, and this
+   pivot energy can not be determined in a fit together with the other
+   model parameters. The reason is that the pivot energy is not an
+   independent parameter of these models, and hence when all other
+   spectral parameters are free, the pivot energy is unconstrained.
+   So please make sure that the pivot energy is fixed, or fix other
+   parameters of the model to assure non-degeneracy of the free
+   parameters.
+