From 000585e98407602de6136f2dfbfac96538b0a8d0 Mon Sep 17 00:00:00 2001 From: Luigi Tibaldo Date: Wed, 7 Nov 2018 10:29:53 +0100 Subject: [PATCH] improved description of SUBDIVSQRT and SIGNIFICANCE algortithms for csresmap and csrespec --- doc/source/users/reference_manual/csresmap.rst | 9 +++++++-- doc/source/users/reference_manual/csresspec.rst | 10 ++++++++-- 2 files changed, 15 insertions(+), 4 deletions(-) diff --git a/doc/source/users/reference_manual/csresmap.rst b/doc/source/users/reference_manual/csresmap.rst index fdd7e45..1877154 100755 --- a/doc/source/users/reference_manual/csresmap.rst +++ b/doc/source/users/reference_manual/csresmap.rst @@ -21,8 +21,11 @@ energies. Four options exist then for residual computation: - the subtraction of the model from the counts (SUB) - the subtraction and division by the model (SUBDIV) - the subtraction and division by the square root of the model (SUBDIVSQRT) -- the computation of the significance using Li & Ma (SIGNIFICANCE) +- the computation of the significance based on the likelihood-ratio + test for Poisson statistics (SIGNIFICANCE) +The SUBDIVSQRT and SIGNIFICANCE algorithms approximate the residual +significance and may become inaccurate in the low-counting regime. The residual map is written into a FITS file. @@ -108,7 +111,9 @@ General parameters ``SUBDIVSQRT``: :math:`(DATA - MODEL) / \sqrt{MODEL}` - ``SIGNIFICANCE``: :math:`{\rm sign}(DATA-MODEL) \times \sqrt{ 2 \times ( DATA \times \ln \left(\frac{DATA}{MODEL} \right) + MODEL - DATA ) }` + ``SIGNIFICANCE``: :math:`{\rm sign}(DATA-MODEL) \times \sqrt{ 2 + \times ( DATA \times \ln \left(\frac{DATA}{MODEL} \right) + + MODEL - DATA ) }` Standard parameters diff --git a/doc/source/users/reference_manual/csresspec.rst b/doc/source/users/reference_manual/csresspec.rst index e08a53b..de91b46 100644 --- a/doc/source/users/reference_manual/csresspec.rst +++ b/doc/source/users/reference_manual/csresspec.rst @@ -35,7 +35,11 @@ Four options exist then for residual computation: - the subtraction of the model from the counts (SUB) - the subtraction and division by the model (SUBDIV) - the subtraction and division by the square root of the model (SUBDIVSQRT) -- the computation of the significance using Li & Ma (SIGNIFICANCE) +- the computation of the significance based on the likelihood-ratio + test for Poisson statistics (SIGNIFICANCE) + +The SUBDIVSQRT and SIGNIFICANCE algorithms approximate the residual +significance and may become inaccurate in the low-counting regime. The counts, model, and residual spectra are written into a FITS file. @@ -162,7 +166,9 @@ General parameters ``SUBDIVSQRT``: :math:`(DATA - MODEL) / \sqrt{MODEL}` - ``SIGNIFICANCE``: :math:`{\rm sign}(DATA-MODEL) \times \sqrt{ 2 \times ( DATA \times \ln \left(\frac{DATA}{MODEL} \right) + MODEL - DATA ) }` + ``SIGNIFICANCE``: :math:`{\rm sign}(DATA-MODEL) \times \sqrt{ 2 + \times ( DATA \times \ln \left(\frac{DATA}{MODEL} \right) + + MODEL - DATA ) }` Standard parameters -- libgit2 0.21.4