Installation

You will need to install the GNU Scientific Library (GSL).

If you use conda, you can install the GSL with:

conda install -c conda-forge pkg-config gsl

If you use Ubuntu, you can use:

sudo apt install libgsl-dev pkg-config

Then, install the CPG likelihood package with

pip install git+https://github.com/ghcollin/cpg_likelihood.git

Downloading

To download, use the following command to clone the repository and all submodules:

git clone --recurse-submodules https://github.com/ghcollin/cpg_likelihood.git

Optional requirements

To use the run_emcee helper function, you will need to install the emcee package:

pip install emcee

To use the run_ptemcee helper function, you will need to install the ptemcee package:

pip install ptemcee

Examples

A file with examples for setting up a model and running the emcee sampler is located in

examples/model_construction.py

A file with examples for using the dynesty sampler and creating the mu(epsilon) function through integration, simulation or from detector Monte-Carlo is located in

examples/mueps_construction.py

You will need to install the emcee, scipy, matplotlib, corner and dynesty packages to run these examples:

pip install emcee scipy matplotlib corner dynesty

You can then run the examples by specifying an output PDF into which the posterior will be rendered:

python3 model_construction.py test.pdf

Current limitations

Only a single break for the differential source count function is currently implemented. Multiple breaks will be added in a future version.

Dependencies

In addition to GSL, this library makes use of the Eigen C++ linear algebra library. A recent development version of Eigen is linked into this repository though a git submodule, as the latest stable version does not include features required by this library. A modified version of the exponential integral algorithm written by Guillermo Navas-Palencia and described in this paper is included in this repository. The algorithm has been modified to make it more numerically stable for this specific application, where catastrophic cancellation can occur when the source flux is small.

References

If you use this library, you may wish to consider citing

@article{navas-palenciaFastAccurateAlgorithm2018,
    title = {Fast and Accurate Algorithm for the Generalized Exponential Integral {{E}} ν (x) for Positive Real Order},
    author = {Navas-Palencia, Guillermo},
    date = {2018-02},
    journal = {Numerical Algorithms},
    shortjournal = {Numer Algor},
    volume = {77},
    pages = {603--630},
    issn = {1017-1398, 1572-9265},
    doi = {10.1007/s11075-017-0331-z},
    url = {http://link.springer.com/10.1007/s11075-017-0331-z},
    urldate = {2020-10-27},
    langid = {english},
    number = {2}
}

@article{Collin:2021ufc,
    title = {A {{Compound Poisson Generator}} Approach to {{Point}}-{{Source Inference}} in {{Astrophysics}}},
    author = {Collin, Gabriel H. and Rodd, Nicholas L. and Erjavec, Tyler and Perez, Kerstin},
    date = {2021-04-09},
    url = {http://arxiv.org/abs/2104.04529},
    urldate = {2021-04-13},
    archivePrefix = {arXiv},
    eprint = {2104.04529},
    eprinttype = {arxiv}
}