Optimized statistical approach for combining multi-messenger data for neutron star equation of state inference. (arXiv:2004.00656v1 [astro-ph.HE])
<a href="http://arxiv.org/find/astro-ph/1/au:+Raithel_C/0/1/0/all/0/1">Carolyn Raithel</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Ozel_F/0/1/0/all/0/1">Feryal Ozel</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Psaltis_D/0/1/0/all/0/1">Dimitrios Psaltis</a>
The neutron star equation of state (EOS) is now being constrained from a
diverse set of multi-messenger data, including gravitational waves from binary
neutron star mergers, X-ray observations of the neutron star radius, and many
types of laboratory nuclear experiments. These measurements are typically
mapped to a common domain — either to a corresponding radius or to a
parametrized EOS using a Bayesian inference scheme — for comparison with one
another. We explore here the statistical biases that can arise when such
multi-messenger data are mapped to a common domain for comparison. We find that
placing Bayesian priors individually in each domain of measurement can
transform to biased constraints in the domain of comparison. Using the first
two binary neutron star mergers as an example, we show that a uniform prior in
the tidal deformability can produce artificial evidence for large radii, which
the data do not support. We present a new prescription for defining Bayesian
priors in any domain of measurement, that will allow for minimally-biased
constraints in the domain of comparison. Finally, using this new prescription,
we provide a status update on multi-messenger EOS constraints on the neutron
star radius.
The neutron star equation of state (EOS) is now being constrained from a
diverse set of multi-messenger data, including gravitational waves from binary
neutron star mergers, X-ray observations of the neutron star radius, and many
types of laboratory nuclear experiments. These measurements are typically
mapped to a common domain — either to a corresponding radius or to a
parametrized EOS using a Bayesian inference scheme — for comparison with one
another. We explore here the statistical biases that can arise when such
multi-messenger data are mapped to a common domain for comparison. We find that
placing Bayesian priors individually in each domain of measurement can
transform to biased constraints in the domain of comparison. Using the first
two binary neutron star mergers as an example, we show that a uniform prior in
the tidal deformability can produce artificial evidence for large radii, which
the data do not support. We present a new prescription for defining Bayesian
priors in any domain of measurement, that will allow for minimally-biased
constraints in the domain of comparison. Finally, using this new prescription,
we provide a status update on multi-messenger EOS constraints on the neutron
star radius.
http://arxiv.org/icons/sfx.gif
