Impact of selection biases on tests of general relativity with gravitational-wave inspirals

Tests of general relativity with gravitational-wave observations from merging compact binaries continue to confirm Einstein's theory of gravity with increasing precision. However, these tests have so far been applied only to signals that were first confidently detected by matched-filter searches assuming general relativity templates. This raises the question of selection biases: What is the largest deviation from general relativity that current searches can detect, and are current constraints on such deviations necessarily narrow because they are based on signals that were detected by templated searches in the first place? In this paper, we estimate the impact of selection effects for tests of the inspiral phase evolution of compact binary signals with a simplified version of the GSTLAL search pipeline. We find that selection biases affect the search for very large values of the deviation parameters, much larger than the constraints implied by the detected signals. Therefore, combined population constraints from confidently detected events are mostly unaffected by selection biases, with the largest effect being a broadening at the similar to 10% level for the -1 PN term. These findings suggest that current population constraints on the inspiral phase are robust without factoring in selection biases. Our study does not rule out a disjoint, undetectable binary population with large deviations from general relativity or stronger selection effects in other tests or search procedures.