The SAMI Galaxy Survey: A statistical approach to an optimal classification of stellar kinematics in galaxy surveys

Large galaxy samples from multiobject integral field spectroscopic (IFS) surveys now allow for a statistical analysis of the z ∼ 0 galaxy population using resolved kinematic measurements. However, the improvement in number statistics comes at a cost, with multiobject IFS survey more severely impacted by the effect of seeing and lower signal-to-noise ratio. We present an analysis of ∼1800 galaxies from the SAMI Galaxy Survey taking into account these effects. We investigate the spread and overlap in the kinematic distributions of the spin parameter proxy λReλRe as a function of stellar mass and ellipticity ee. For SAMI data, the distributions of galaxies identified as regular and non-regular rotators with KINEMETRY show considerable overlap in the λReλRe–ee diagram. In contrast, visually classified galaxies (obvious and non-obvious rotators) are better separated in λReλRe space, with less overlap of both distributions. Then, we use a Bayesian mixture model to analyse the observed λReλRe–log (M⋆/M⊙) distribution. By allowing the mixture probability to vary as a function of mass, we investigate whether the data are best fit with a single kinematic distribution or with two. Below log (M⋆/M⊙) ∼ 10.5, a single beta distribution is sufficient to fit the complete λReλRe distribution, whereas a second beta distribution is required above log (M⋆/M⊙) ∼ 10.5 to account for a population of low-λReλRe galaxies. While the Bayesian mixture model presents the cleanest separation of the two kinematic populations, we find the unique information provided by visual classification of galaxy kinematic maps should not be disregarded in future studies. Applied to mock-observations from different cosmological simulations, the mixture model also predicts bimodal λReλRe distributions, albeit with different positions of the λReλRe peaks. Our analysis validates the conclusions from previous, smaller IFS surveys, but also demonstrates the importance of using selection criteria for identifying different kinematic classes that are dictated by the quality and resolution of the observed or simulated data.