Data Selection Criteria for Spectroscopic Measurements of Neutron Star Radii with X-ray Bursts

TL;DR

This paper compares data-driven and theory-based methods for selecting X-ray burst data to measure neutron star radii, finding that data-driven methods are more practical given current data limitations.

Contribution

It demonstrates that theoretical selection criteria are impractical with current data and shows that data-driven approaches are more robust for spectroscopic measurements.

Findings

Theoretically expected trends are not discernible in most data.

Theoretical selection criteria often select inconsistent data subsets.

Data-driven methods are less limited and more reliable.

Abstract

Data selection and the determination of systematic uncertainties in the spectroscopic measurements of neutron star radii from thermonuclear X-ray bursts have been the subject of numerous recent studies. In one approach, the uncertainties and outliers were determined by a data-driven Bayesian mixture model, whereas in a second approach, data selection was performed by requiring that the observations follow theoretical expectations. We show here that, due to inherent limitations in the data, the theoretically expected trends are not discernible in the majority of X-ray bursts even if they are present. Therefore, the proposed theoretical selection criteria are not practical with the current data for distinguishing clean data sets from outliers. Furthermore, when the data limitations are not taken into account, the theoretically motivated approach selects a small subset of bursts with properties that are in fact inconsistent with the underlying assumptions of the method. We conclude that the data-driven selection methods do not suffer from the limitations of this theoretically motivated one.