This project will develop a quantitative statistical technique to assess the pixel-level fidelity of radio-interferometric images, which are formed by solving a complex calibration and inverse imaging problem. Recent theoretical advances in statistical resampling techniques for dependent data, and geometric increases in high performance computing resources, allow an innovative approach using modern bootstrap methods. This research will incorporate more complete statistical techniques, derive quantitative guidelines for the use of this method, and explore the theoretical properties of resampling. If successful, these methods can be applied to any astronomical interferometry, as well as to other inverse imaging problems in the physical sciences.
Because calibration and imaging requires specialized technical knowledge, which acts as an entry barrier for new users, the development of robust statistical techniques is a vital ingredient required to build automated pipeline reduction systems capable of producing science products with quantitative uncertainties ready for direct use by a far broader user community. This work will therefore be an important contribution to research infrastructure.
