octofitterpy

octofitterpy is a python package for performing Bayesian inference against a wide variety of exoplanet / binary star data. It uses the Octofitter.jl julia package under the hood (just like eg numpy uses C).

octofitterpy can access almost all functionality of Octofitter.jl. Currently a subset of this functionality including relative astrometry fitting, absolute astrometry fitting, and various plotting functions have been wrapped with convenient python functions. Remaining functionality can be accessed via the octofitterpy.Octofitter submodule.

The examples directory and demo notebook provide an introduction to using octofitter in Python. Extensive documentation and tutorials are available here for the Julia version, and for the most part are directly translatable to Python.

Installation

In python 3.8 to 3.11 based environment, run:

pip install -U octofitterpy

Example

import octofitterpy as octo

# See demo.ipynb for more details
astrom_like = octo.PlanetRelAstromLikelihood(
    epoch = [50000,50120],
    sep = [505.7,600.1],
    pa = [0.0,0.4,],
    σ_sep = [10,10],
    σ_pa = [0.01,0.01],
    cor= [0,0.2]
)
planet_b = octo.Planet(
    name="b",
    basis="Visual{KepOrbit}",
    priors=
    """            
        a ~ LogUniform(0.1, 500)
        e ~ Uniform(0.0, 0.99)
        i ~ Sine()
        ω ~ UniformCircular()
        Ω ~ UniformCircular()
        θ ~ UniformCircular()
        tp = θ_at_epoch_to_tperi(system,b,50000) # use MJD epoch of your data here!!
    """,
    likelihoods=[astrom_like]
)
sys = octo.System(
    name="HIP100123",
    priors = 
    """
        M ~ truncated(Normal(1.2, 0.1), lower=0)
        plx ~ truncated(Normal(50.0, 0.02), lower=0)
    """,
    likelihoods=[],
    companions=[planet_b]
)
model = octo.LogDensityModel(sys) # Compile model
chain = octo.octofit(model) # Sample model
octo.octoplot(model,chain) # Plot orbits
octo.octocorner(model,chain,small=True) # Make corner plot
octo.savechain("table.fits", chain)

Read the paper

In addition to these documentation and tutorial pages, you can read the paper published in the Astronomical Journal (open-access).

Attribution

If you use Octofitter in your work, please cite Thompson et al:

@article{Thompson_2023,
    doi = {10.3847/1538-3881/acf5cc},
    url = {https://dx.doi.org/10.3847/1538-3881/acf5cc},
    year = {2023},
    month = {sep},
    publisher = {The American Astronomical Society},
    volume = {166},
    number = {4},
    pages = {164},
    author = {William Thompson and Jensen Lawrence and Dori Blakely and Christian Marois and Jason Wang and Mosé Giordano and Timothy Brandt and Doug Johnstone and Jean-Baptiste Ruffio and S. Mark Ammons and Katie A. Crotts and Clarissa R. Do Ó and Eileen C. Gonzales and Malena Rice},
    title = {Octofitter: Fast, Flexible, and Accurate Orbit Modeling to Detect Exoplanets},
    journal = {The Astronomical Journal},
}

If you use the pairplot functionality, please cite:

@misc{Thompson2023,
    author = {William Thompson},
    title = {{PairPlots.jl} Beautiful and flexible visualizations of high dimensional data},
    year = {2023},
    howpublished = {\url{https://sefffal.github.io/PairPlots.jl/dev}},
}

The python wrapper octofitterpy is based on the excellent PySR by Miles Cranmer.
See the documentation for a list of additional papers to consider citing.