BANYAN II web tool

by Jonathan Gagné

BANYAN II: Bayesian Analysis for Nearby Young AssociatioNs II

Membership probability without photometric information

(Version 1.4)


BANYAN II is a Bayesian analysis tool to determine the membership probability of candidate stars to nearby young kinematic groups. This pipeline is based on a comparison of Galactic position (XYZ) and space velocity (UVW), using a naive Bayesian classifier. We only consider well-defined moving groups closer than 100 pc and younger than 200 Myr. When radial velocity and/or distance measurements are not available, Bayesian inference has the advantage of being able to marginalize them and compute membership probabilities without them. When this is the case, BANYAN II produces statistical predictions for these quantities, according to each membership hypothesis. When a given hypothesis is respected, the agreement between trigonometric distances and BANYAN II statistical distances generally agree within 8%, whereas predicted and measured radial velocities agree within 1.6 km/s.

All modifications included in this BANYAN II web tool with respect to the BANYAN I web tool are listed here :

  • The spatial and kinematic models of moving groups are modeled with 3D gaussian ellipsoids in both BANYAN I and II, but in BANYAN II those ellipsoids are free to have axes oriented in any direction, rather than being constrained in the local Galactic coordinates (XYZ) directions.

  • The list of bona fide members used in the construction of the spatial and kinematic models was updated as of December 2013.

  • The field hypothesis is modeled using a Besançon Galactic model (Robin et al. 2012).

  • Prior probabilities are not set to unity anymore, but rather to the expected populations in each hypothesis considered, to bring the Bayesian probabilities closer to absolute values. This treatment of relative populations takes into account that objects with A) smaller galactic latitude, b) smaller proper motion, c) smaller RVs and d) larger distances, are more likely to be field contaminants.

  • The likelihood functions that are input in Baye's theorem are functions of X,Y,Z,U,V,W instead of the direct observable quantities (ra, dec, pmra, pmdec, RV, distance). The reason for this preference is that former quantities are better represented by gaussian likelihood probability density functions.

  • Prior likelihoods associated to marginalized parameters (distance and/or RV) are computed directly from a Montecarlo drawing of synthetic objects from each association's spatial and kinematic model, which means that they can have any functional form (generally not gaussians).


  • The details of this method are described in Gagné et al. (2014ApJ...783..121G). Please reference this work when you make use of this tool, as well as Malo et al. (2013).


    We also encourage trying David Rodriguez' convergent point analysis online tool.

    We would like to thank Michael Liu for useful comments on this web page.

    Version history.

    v1.0 22/12/2013
    v1.1 24/12/2013 Probabilities taking into account only a RV or PLX measurement (but not both) were mistaken, due to a misdistribution of prior probabilities (every hypothesis always inherited the Beta Pictoris prior probability - depending on some cases, this could have a very small or very big effect on the probabilities). This has been fixed from this version, as well as in the v2+ of the arxiv version of Gagné et al. (2014). The ApJ version was unaffected.
    v1.2 3/1/2014 Probabilities taking into account either a RV or PLX measurement (or both) were slightly offset, because the ξν (related to RV) and ξϖ (related to distance) were always = 1 (See Section 3.1 of Gagné et al. 2014 for more info). This has been fixed from this version. This error was not present in the paper, and thus did not affect either the arxiv or ApJ version.
    v1.3 23/4/2014 RV and PLX entries without decimals were previously (mistakenly) treated as integers, hence yielding problematic results.
    v1.4 11/8/2015 Separated the output probabilities of "Young" and "Old" field, added the statistical RVs for the field hypotheses, and fixed glitches in the output format of statistical RVs and distances.

    The results of this tool should be interpreted with caution.

    1) We stress that a high probability in one of the associations considered here does not necessarily imply that the candidate star is young. Before an object is considered as a new bona fide member to a moving group, it must show signs of youth as well as spatial and kinematic agreement with the moving group in question (using both a RV and parallax measurements). Conversely, a high probability in the « field » hypothesis does not necessarily imply that the star is old; it simply means that it is less likely to be a member of the young associations considered here compared to the « field »

    2) Users must keep in mind that Bayesian probabilities might be wrong when studying members of moving groups or associations not considered in our hypotheses. For example, a member of Upper Scorpius could come up as a good candidate of AB Doradus, because the models used by our tool do not consider the former.

    3) Bayesian probabilities reported by BANYAN II will often be lower than those reported by BANYAN I, because the priors were not set to unity here, which strongly favors the field that generally has a significantly higher expected population. This means in turn that any object with Bayesian probabilities not negligible for a given moving group is worthy of further study (e.g. a 20% probability in BANYAN II could amount to a 90% probability in BANYAN I).

    4) Bayesian probabilities calculated using a naive Bayesian classifier (which is the case in BANYAN I and II) are subject to being biased because of the independent treatment of observables that are not independent in reality. For a given object, this will not affect the relative classification rank of hypotheses (e.g. from least probable to most probable), but it will affect the absolute values of probabilities (see Hand and Yu 2001).

    5) Results obtained here may differ from those published in Gagné et al. (2014ApJ...783..121G) as the latter includes 2MASS and WISE photometry in computing membership probabilities.

    6) The information entered on this web page is not stored on our server.


    Instructions :

  • You can simply enter the name of a star, as resolved by Simbad or Vizier.

  • Press the RESOLVE button. BANYAN II will seek all the information it can find from a limited number of online catalogs. Missing information will be reported as « NaN ».Be patient, this may take several tens of seconds.

  • You can modify and/or remove some information as you please but you need minimally a sky position, proper motion and a measurement error on proper motion to proceed.

  • Proper motions are in units of mas/yr, and the proper motion in the RA direction is implicitly understood as the usual mu_ra * cos(delta), where delta is the declination of the star.

  • Press SUBMIT and read the cautionary note above (if not already done) before interpreting the results.

  • You can use the "set priors to unity" checkbox in order to compare probabilities with BANYAN I. However, these probabilities will be more strongly biased (far from absolute probabilities ; See Gagné et al. 2014.)

  • The main effect of checking "Object is younger than 1 Gyr" is that prior probabilities are different (apart from the fact that the field population dymanics is dependent on age) - this is so because the expected population of old field objects is significantly larger than that of young objects. Hence, this option will generally have a very small effect on Bayesian probabilities if the "set priors to unity" option is checked.

  • If you check the "Younger than 1 Gyr" option, the "Old field" hypothesis is withdrawn from the analysis, and hence its associated probability is automatically zero. In that case, all the reported probabilities, statistical distances and statistical RVs are for the young field. If you do not check the box, separate probabilities, RV and distance predictions will be given out - "OFLD" will correspond to the old field, and "YFLD" to the young field.

  • Please be reminded that this tool does not use photometry as input observables. Most of the papers about brown dwarfs in the BANYAN series use an augmented version of BANYAN II that does use 2MASS and WISE photometry, hence the resulting output probabilities will be different.

  • Acknowledgement : if your paper uses results obtained with BANYAN II, please kindly give a reference to Gagné et al. (2014ApJ...783..121G) and Malo et al. (2013).

  • If you have any questions and/or comments, contact me at gagne (@) astro (dot) umontreal (dot) ca.

  • NAME of your star: PRESS:

    Check this box if this object is younger than 1 Gyr.
    Set priors to unity.

    Right ascension (degrees) :

    Proper motion in right ascension (mas/yr) :

    Proper motion in declination (mas/yr) :

    Radial velocity (km/s) :

    Parallax (mas) :

    STEP 2 :






    Declination (degrees) :

    Error on Proper motion in right ascension (mas/yr) :

    Error on Proper motion in declination (mas/yr) :

    Error on radial velocity (km/s) :

    Error on parallax (mas) :

    Name of your star:








    Observational properties (UVWXYZ) of the young associations considered in the analysis :
    *Note that primed quantities are in measured in a rotated frame of reference (see Gagne et al. 2014) :
    Name Age (Myr) 1-σ distance range (pc) 1-σ RV range (km/s) U (km/s) V (km/s) W (km/s) σ U' (km/s) σ V' (km/s) σ W' (km/s) X (pc) Y (pc) Z (pc) σ X' (pc) σ Y' (pc) σ Z' (pc) # stars
    TW Hydrae 8 - 12 40 - 62 (7) - (12) -11.1 -18.9 -5.6 0.9 1.6 2.8 19.1 -54.2 21.5 5.0 7.2 22.6 18
    β Pictoris 12 - 22 18 - 40 (-9) - (16) -11.0 -15.6 -9.2 1.4 1.7 2.5 7.6 -3.5 -14.5 8.2 13.5 30.7 33
    Tucana-Horologium 20 - 40 38 - 51 (3) - (14) -9.7 -20.5 -0.8 1.1 1.7 2.4 6.7 -21.8 -36.0 3.9 10.6 20.1 52
    Columba 20 - 40 26 - 63 (19) - (26) -12.1 -21.3 -5.6 0.5 1.3 1.7 -28.1 -25.8 -28.6 10.5 17.6 28.3 21
    Carina 20 - 40 11 - 42 (16) - (23) -10.7 -22.2 -5.7 0.3 0.7 1.1 10.1 -51.6 -14.9 5.8 11.3 29.8 21
    Argus 30 - 50 15 - 48 (-10) - (9) -21.5 -12.2 -4.6 0.9 1.7 2.7 15.0 -21.7 -8.1 12.1 15.5 27.4 11
    AB Doradus 70 - 120 19 - 50 (-11) - (29) -7.0 -27.2 -13.9 1.2 1.7 1.9 -2.5 1.3 -16.3 16.3 20.0 23.5 54
    Young Field < 1000 66 - 169 (-19) - (19) -11.2 -18.6 -6.9 7.7 12.5 19.6 2.8 0.1 -13.1 79.6 80.4 80.8 -
    Old Field > 1000 70 - 177 (-34) - (32) -11.2 -18.6 -6.9 7.7 12.5 19.6 2.8 0.1 -13.1 79.6 80.4 80.8 -