Observational selection biases in time-delay strong lensing and their impact on cosmography
Inferring cosmological parameters from time-delay strong lenses requires a significant investment of telescope time; it is therefore tempting to focus on the systems with the brightest sources, the highest image multiplicities and the widest image separations. We investigate if this selection bias can influence the properties of the lenses studied and the cosmological parameters that are inferred. Using a population of lenses with ellipsoidal powerlaw density profiles, we build a sample of double and quadruple image systems. Assuming reasonable thresholds on image separation and flux, based on current lens monitoring campaigns, we find that the typical density profile slopes of monitorable lenses are significantly shallower than the input ensemble. From a sample of quadruple image lenses we find that this selection function can introduce a 3.5% bias on the inferred time-delay distances if the ensemble of deflector properties is used as a prior for a cosmographical analysis. This bias remains at the 2.4% level when high resolution imaging of the quasar host is used to precisely infer the density profiles of individual lenses. We also investigate if the lines-of-sight for monitorable strong lenses are biased. After adding external convergence, $κ$, and shear to our lens population we find that the expectation value for $κ$ is increased by 0.004 and 0.009 for doubles and quads respectively. $κ$ is degenerate with the value of $H_0$ inferred from time delays; fortunately the shift in $κ$ only induces a 0.9 (0.4) percent bias on $H_0$ for quads (doubles). We therefore conclude that whilst the properties of typical quasar lenses and their lines-of-sight do deviate from the global population, the total magnitude of this effect is likely a subdominant effect for current analyses, but has the potential to be a major systematic for samples of $\sim$25 or more lenses.