Multidimensional hierarchical tests of general relativity with gravitational waves

Tests of general relativity with gravitational waves typically introduce parameters for putative deviations and combine information from multiple events by characterizing the population distribution of these parameters through a hierarchical model. Although many tests include multiple such parameters, hierarchical tests have so far been unable to accommodate this multidimensionality, instead restricting to separate one-dimensional analyses and discarding information about parameter correlations. In this paper, we extend population tests of general relativity to handle an arbitrary number of dimensions. We demonstrate this framework on the two-dimensional inspiral-merger-ringdown consistency test, and derive new constraints from the latest LIGO-Virgo-KAGRA catalog, GWTC-3. We obtain joint constraints for the two parameters introduced by the classic formulation of this test, revealing their correlation structure both at the individual-event and population levels. We additionally propose a new four-dimensional formulation of the inspiral-merger-ringdown test that we show contains further information. As in past work, we find the GW190814 event to be an outlier; the 4D analysis yields further insights on how the low mass and spin of this event biases the population results. Without (with) this event, we find consistency with general relativity at the 60% (92%) credible level in the 2D formulation, and 76% (80%) for the 4D formulation. This multidimensional framework can be immediately applied to other tests of general relativity in any number of dimensions, including the parametrized post-Einsteinian tests and ringdown tests.