What if we find nothing? Bayesian analysis of the statistical information of null results in future exoplanet habitability and biosignature surveys

Future telescopes will survey temperate, terrestrial exoplanets to estimate the frequency of habitable (${\eta }_{\text{Hab}}$) or inhabited (${\eta }_{\text{Life}}$) planets. This study aims to determine the minimum number of planets ($N$) required to draw statistically significant conclusions, particularly in the case of a null result (i.e., no detections). Using a Bayesian framework, we analyzed surveys of up to $N=100$ planets to infer the frequency of a binary observable feature (${\eta }_{\text{obs}}$) after null results. Posterior best fits and upper limits were derived for various survey sizes and compared with predicted yields from missions like the Large Interferometer for Exoplanets (LIFE) and the Habitable Worlds Observatory (HWO). Our findings indicate that $N=20-50$ perfect'' observations (100\% confidence in detecting or excluding the feature) yield conclusions relatively independent of priors. To achieve 99.9\% upper limits of $η_{\text{obs}} \leq 0.2/0.1$, approximately $N \simeq 40/80$ observations are needed. For imperfect’’ observations, uncertainties in interpretation and sample biases become limiting factors. We show that LIFE and HWO aim for sufficiently large survey sizes to provide statistically meaningful estimates of habitable environments and life prevalence under these assumptions. However, robust conclusions require careful sample selection and high-confidence detection or exclusion of features in each observation.