Bootstrap resampling as a tool for radio-interferometric imaging fidelity assessment

Abstract

We report on a numerical evaluation of the statistical bootstrap as a technique for radio-interferometric imaging fidelity assessment. The development of a fidelity assessment technique is an important scientific prerequisite for automated pipeline reduction of data from modern radio interferometers. We evaluate the statistical performance of two bootstrap methods, the model-based bootstrap and subsample bootstrap, against a Monte Carlo parametric simulation, using interferometric polarization calibration and imaging as the representative problem under study. We find both statistical resampling techniques to be viable approaches to radio-interferometric imaging fidelity assessment which merit further investigation. We also report on the development and implementation of a new self-calibration algorithm for radio-interferometric polarimetry which makes no approximations for the polarization source model.