phoetting: guidelines and best practices for fitting with PHOEBE

The latest release of PHOEBE (PHysics Of Eclipsing BinariEs) features extensive support for fitting methods, including estimators based on the geometry of the light and radial velocity curves, an interface for several scipy.optimize methods, as well as wrappers around emcee and dynesty for MCMC and nested sampling, respectively. These, alongside the support for alternate backends (ellc, jktebop, WD) make fitting eclipsing binary data with PHOEBE exceptionally straightforward. However, each eclipsing binary is unique and more often than not requires careful analysis and consideration of the scientific and numerical methods utilized in its analysis.

Here we summarize some common pitfalls of fitting with PHOEBE that arise due to the underlying complex model and how it interacts with the fitting methods. We discuss the reliability of estimators in terms of light curve morphology and phase coverage; different choices of parameterization and its effects on optimization and sampling; how the use of atmosphere tables can skew the values of certain parameters; marginalization over nuisance parameters versus a-priori assumptions, as well as the effect of data uncertainties on the end result. We offer guidelines and best practices that ensure the most scientifically reliable results.