Abstract: The multi-detector F-statistic is close to optimal for detecting continuous gravitational waves (CWs) in Gaussian noise. However, it is susceptible to false alarms from instrumental artefacts, for example quasi-monochromatic disturbances ('lines'), which resemble a CW signal more than Gaussian noise. In a recent paper [Keitel et al 2014, PRD 89 064023], a Bayesian model selection approach was used to derive line-robust detection statistics for CW signals, generalising both the F-statistic and the F-statistic consistency veto technique and yielding improved performance in line-affected data. Here we investigate a generalisation of the assumptions made in that paper: if a CW analysis uses data from two or more detectors with very different sensitivities, the line-robust statistics could be less effective. We investigate the boundaries within which they are still safe to use, in comparison with the F-statistic. Tests using synthetic draws show that the optimally-tuned version of the original line-robust statistic remains safe in most cases of practical interest. We also explore a simple idea on further improving the detection power and safety of these statistics, which we however find to be of limited practical use.
