From 65bcb4a1f4064dd13dffa06495713d5229b4a8bd Mon Sep 17 00:00:00 2001 From: Jürgen Knödlseder Date: Fri, 30 Oct 2015 22:38:34 +0100 Subject: [PATCH] Add some more known issues related to El Capitan and the elliptical Gaussian model. --- doc/source/issues.rst | 79 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------- 1 file changed, 66 insertions(+), 13 deletions(-) diff --git a/doc/source/issues.rst b/doc/source/issues.rst index fe2f1fb..a752f2e 100644 --- a/doc/source/issues.rst +++ b/doc/source/issues.rst @@ -5,11 +5,68 @@ Known issues Below you will find a list of known issues. +**Installation issues** + +- :ref:`Certificate problem when cloning from Git ` +- :ref:`Python unit test fail on El Capitan ` + +**Analysis issues** + +- :ref:`Binned analysis is biased for coarse binning ` + +**Model fitting issues** + - :ref:`Broken power law has unreliable errors ` - :ref:`Errors become unreliable when fitting the pivot energy ` -- :ref:`Binned analysis is biased when using coarse binning ` - :ref:`Shell model is biased when width is comparable to angular resolution ` +- :ref:`Elliptical Gaussian model fit converges slowly ` + + +Installation issues +^^^^^^^^^^^^^^^^^^^ + +.. _issue_ssl: + +.. topic:: Certificate problem when cloning from Git + + When cloning ctools from git you may encounter an SSL certificate + problem. This is related to the usage of the https protocol and can + be resolved by setting the ``GIT_SSL_NO_VERIFY`` environment variable + to true: + + .. code-block:: bash + export GIT_SSL_NO_VERIFY=true + +.. _issue_elcapitan: + +.. topic:: Python unit test fail on El Capitan + + On Mac OS X 10.11 (El Capitan) the Python unit tests fail when ctools + are not installed due to the way the ``DYLD_LIBRARY_PATH`` environment + variable is handled by the operating system. A (not very elegant) + workaround is to install the ctools before running the unit test using + ``make check``. + + +Analysis issues +^^^^^^^^^^^^^^^ + +.. _issue_binned: + +.. topic:: Binned analysis is biased for coarse binning + + When performing a binned or stacked analysis you should make sure + that the spatial and spectral binning is sufficiently fine grained. + The spatial binning should be better than the best angular resolution + over the energy range of interest. Use a typical value of 0.02 degrees + per pixel for the spatial binning and 10 bins per decade for the + spectral binning. If the binning is too coarse, the spectral parameters + that are fitted will be biased. + + +Model fitting issues +^^^^^^^^^^^^^^^^^^^^ .. _issue_bplaw: @@ -35,18 +92,6 @@ Below you will find a list of known issues. parameters of the model to assure non-degeneracy of the free parameters. -.. _issue_binned: - -.. topic:: Binned analysis is biased when using coarse binning - - When performing a binned or stacked analysis you should make sure - that the spatial and spectral binning is sufficiently fine grained. - The spatial binning should be better than the best angular resolution - over the energy range of interest. Use a typical value of 0.02 degrees - per pixel for the spatial binning and 10 bins per decade for the - spectral binning. If the binning is too coarse, the spectral parameters - that are fitted will be biased. - .. _issue_shell: .. topic:: Shell model is biased when width is comparable to angular resolution @@ -57,3 +102,11 @@ Below you will find a list of known issues. The fitted shell width and radius should thus not be overinterpreted when the width is close to the angular resolution of CTA. +.. _issue_egauss: + +.. topic:: Elliptical Gaussian model fit converges slowly + + The convergence of the elliptical Gaussian model can be slow and + in some situations requires of the order of 20 iterations before + the fit terminates. Nevertheless, the numerical accuracy of the model + fitting results are satisfactory. -- libgit2 0.21.4