A significance test for mixed-integer models with application to GNSS

Abstract GNSS model validation constitutes an essential part of any GNSS data processing scheme. With the inclusion of the very precise, but integer ambiguous carrier-phase data, the GNSS models become of the mixed-integer type. Although inference theory of mixed-integer models is well developed for parameter estimation, this is not yet the case for the validity testing of such models. It is the goal of this contribution to help close this gap by introducing the ambiguity-resolved parameter significance test. It differs from existing significance tests in that it takes the unknown integerness of the ambiguities rigorously into account. Our analysis shows that the proposed test can significantly outperform currently used tests.