Unbiased inference of the masses of transiting planets from radial velocity followup

TL;DR

This paper highlights how stopping radial velocity observations based on preliminary mass estimates biases the inferred planet masses upward, and recommends pre-planned, unbiased observing strategies to ensure accurate planetary characterization.

Contribution

It demonstrates the bias introduced by adaptive stopping criteria in RV followup and advocates for pre-defined, unbiased observation plans to improve mass measurements.

Findings

Bias in mass estimates increases with adaptive stopping.

Pre-planned observation strategies reduce mass measurement bias.

Publishing non-detections helps avoid biased inferences.

Abstract

Data from the TESS mission will be used to discover hundreds of small planets amenable to radial velocity (RV) followup. Often, RV observations are obtained until a particular fractional precision on the inferred mass is achieved. I show that when the decision to stop collecting RV observations depends on the mass inferred from already-collected data, this will bias mass measurements upward, particularly when the fixed length of RV observing campaigns is considered. I suggest that observing teams should determine their observing strategy for each star before any data are collected, and all stopping criteria should not directly depend on the inferred mass. Observing teams should explicitly publish both their criteria for observing targets and deciding to end their observations, as well as their mass non-detections to avoid introducing biases into the masses---and thus inferences on the densities, compositions, and atmospheres---of transiting planets.