We investigate potential biases in the measurements of exoplanet orbital parameters obtained from radial velocity observations for single-planet systems .
We create a mock catalog of radial velocity data , choosing input planet masses , periods , and observing patterns from actual radial velocity surveys and varying input eccentricities .
We apply Markov Chain Monte Carlo ( MCMC ) simulations and compare the resulting orbital parameters to the input values .
We find that a combination of the effective signal-to-noise ratio of the data , the maximal gap in phase coverage , and the total number of periods covered by observations is a good predictor of the quality of derived orbit parameters .
As eccentricity is positive definite , we find that eccentricities of planets on nearly circular orbits are preferentially overestimated , with typical bias of 1 - 2 times the median eccentricity uncertainty in a survey ( e.g. , 0.04 in the Butler et al . 4 catalog ) .
When performing population analysis , we recommend using the mode of the marginalized posterior eccentricity distribution to minimize potential biases .
While the Butler et al .
( 4 ) catalog reports eccentricities below 0.05 for just 17 % of single-planet systems , we estimate that the true fraction of e \leq 0.05 orbits is about f _ { 0.05 } = 38 \pm 9 % .
For planets with P > 10 days , we find f _ { 0.05 } = 28 \pm 8 % versus 10 % from Butler et al .
( 4 ) .
These planets either never acquired a large eccentricity or were circularized following any significant eccentricity excitation .