2025A&A...697A..25F

The detection of gravitational waves with pulsar timing arrays (PTAs) requires precise measurements of the difference between the pulsars' timing models and their observed pulses as well as dealing with numerous and sometimes hard-to-diagnose sources of noise. Outliers may have an impact on this already difficult procedure, especially if the methods used are not robust to such anomalous observations. Until now, no complete and practical quantification of their effects on PTA data has been provided. With this work, we aim to fill this gap. We corrupted simulated datasets featuring an increasing degree of complexity with varying percentages of uniformly distributed outliers and investigated the impact of the latter on the recovery of the injected gravitational wave signals and pulsar noise terms. We find that the gravitational wave signal, due to its expected correlation, is more robust against these anomalous observations when compared to the other injected processes. This result is especially relevant in the context of the emerging statistical evidence for the gravitational wave background in PTA datasets, further strengthening those claims.