In its prime state, DATA is a Latin word meaning “[things] given”, a plural noun derived from the verb “To Give”. Its singular form is DATUM. Modern conversation equates DATA with “Information”, while modern philosophies on information management are getting entwined with parallel philosophies on knowledge management. In some ways that is a positive development, and is greatly assisted by Open Access and Internet policies, but in others it is more detrimental, by threatening to blur the essential distinction between objectivity and subjectivity in our science. We examine that essential distinction from the view-points of observers, authors (and publishers), and database managers, and suggest where, when and how the distinctiveness of their fundamental contributions to the communication and validation of research results should be respected and upheld.

The IAU's Statistics and Exoplanets Focus Meeting brings together observers, modelers and methodologists to discuss the intricate challenges of extracting and interpreting faint planetary signals from dominant starlight. Initiated by the IAU's new groups concentrating on astroinformatics and astrostatistics, the meeting stimulated the wider exoplanetary community as well as experts in data and science analysis. This proceedings presented selected papers from the Focus Meeting.

The light curves of spotted, rotating stars are often non-sinusoidal and Quasi-Periodic (QP) and a strictly periodic sinusoid is therefore not a representative generative model. Ideally, a physical model of the stellar surface would be conditioned on the data, however the parameters of such models can be highly degenerate.

Precise radial velocity measurements of a star allow to search for planets. But this method has to face with irregularly time series. Stellar variabilities: pulsation, granulation, stellar activity on a short and long timescale, also modify the measure of the radial velocities. There is indeed a growing literature of controversies on how a signal is interpreted as a planet or due to stellar activity. I present how the star variations change the measured RVs, which techniques and indices are used by several teams to disentangle activity and planets, and the future options that are being studied.

We challenge the assumptions present in previous efforts to model the ensemble of detected Kepler systems, which require a dichotomous stellar population of ‘fertile’ and ‘sterile’ planet producing stars. We remove the assumption of Rayleigh distributed mutual inclinations between planets and show that the need for two distinct stellar populations disappears when the inner part of planetary disks are assumed to be flat, rather than flared.

The Kepler mission has produced tables of exoplanet candidates (“KOI table”), as well as tables of transit detections (“TCE table”), hosted at the Exoplanet Archive (http://exoplanetarchive.ipac.caltech.edu). Transit detections in the TCE table that are plausibly due to a transiting object are selected for inclusion in the KOI table. KOI table entries that have not been identified as false positives (FPs) or false alarms (FAs) are classified as planet candidates (PCs, Mullally et al. 2015). A subset of PCs have been confirmed as planetary transits with greater than 99% probability, but most PCs have <99% probability of being true planets. The fraction of PCs that are true transiting planets is the PC reliability rate. The overall PC population is believed to have a reliability rate >90% (Morton & Johnson 2011).

We developed, based on the CoRoT experience, an automated tool called BART for Bayesian Analysis for the Ranking of Transit in order to perform a homogeneous and automated ranking of planetary candidates. We applied it to the candidates detected in the campaign 1 of K2.

Transiting candidate validation is essentially a Bayesian model comparison problem: different models, all explaining the observations comparably well, compete for the support of the available data. The basic characteristics of the planet validation problem are discussed and the different approaches taken to tackle its difficulties are reviewed.

Diverse variable phenomena in the Universe are periodic. Astonishingly many of the periodic signals present in stars have timescales coinciding with human ones (from minutes to years). The periods of signals often have to be deduced from time series which are irregularly sampled and sparse, furthermore correlations between the brightness measurements and their estimated uncertainties are common. The uncertainty on the frequency estimation is reviewed. We explore the astronomical and statistical literature, in both cases of regular and irregular samplings. The frequency uncertainty is depending on signal to noise ratio, the frequency, the observational timespan. The shape of the light curve should also intervene, since sharp features such as exoplanet transits, stellar eclipses, raising branches of pulsation stars give stringent constraints. We propose several procedures (parametric and nonparametric) to estimate the uncertainty on the frequency which are subsequently tested against simulated data to assess their performances.

Time-series photometry and spectroscopy of transiting exoplanets allow us to study their atmospheres. Unfortunately, the required precision to extract atmospheric information surpasses the design specifications of most general purpose instrumentation. This results in instrumental systematics in the light curves that are typically larger than the target precision. Systematics must therefore be modelled, leaving the inference of light-curve parameters conditioned on the subjective choice of systematics models and model-selection criteria. Here, I briefly review the use of systematics models commonly used for transmission and emission spectroscopy, including model selection, marginalisation over models, and stochastic processes. These form a hierarchy of models with increasing degree of objectivity. I argue that marginalisation over many systematics models is a minimal requirement for robust inference. Stochastic models provide even more flexibility and objectivity, and therefore produce the most reliable results. However, no systematics models are perfect, and the best strategy is to compare multiple methods and repeat observations where possible.

A new apodized Keplerian model is proposed for the analysis of precision radial velocity (RV) data to model both planetary and stellar activity (SA) induced RV signals. A symmetrical Gaussian apodization function with unknown width and center can distinguish planetary signals from SA signals on the basis of the width of the apodization function. The general model for m apodized Keplerian signals also includes a linear regression term between RV and the stellar activity diagnostic In (R'hk), as well as an extra Gaussian noise term with unknown standard deviation. The model parameters are explored using a Bayesian fusion MCMC code. A differential version of the Generalized Lomb-Scargle periodogram provides an additional way of distinguishing SA signals and helps guide the choice of new periods. Sample results are reported for a recent international RV blind challenge which included multiple state of the art simulated data sets supported by a variety of stellar activity diagnostics.

Measuring the masses of rocky planets is quite difficult, as the relevant signal produced by such planets is often dwarfed by stellar activity by an order of magnitude or more. Developing a more robust way to isolate the stellar activity in these measurements is crucial to the search for Earth-like planets. We estimate the mass of Earth-size planet Kepler-78b using a Gaussian process estimator to describe the stellar activity in both photometric and radial velocity (RV) data, confirming previous results with a more robust technique that can be extended toward Earth analogues.

Catastrophic failure is an unsolved problem existing in the most photometric redshift estimation approaches for a long history. In this study, we propose a novel approach by integration of k-nearest-neighbor (KNN) and support vector machine (SVM) methods together. Experiments based on the quasar sample from SDSS show that the fusion approach can significantly mitigate catastrophic failure and improve the accuracy of photometric redshift estimation.

Twenty years ago, no planets were known outside our own solar system. Since then, the discoveries of ~1500 exoplanets have radically altered our views of planets and planetary systems. This revolution is due in no small part to the Kepler Mission, which has discovered >1000 of these planets and >4000 planet candidates. While Kepler has shown that small rocky planets and planetary systems are quite common, the quest to find Earth's closest cousins and characterize their atmospheres presses forward with missions such as NASA Explorer Program's Transiting Exoplanet Survey Satellite (TESS) slated for launch in 2017 and ESA's PLATO mission scheduled for launch in 2024.

These future missions pose daunting data processing challenges in terms of the number of stars, the amount of data, and the difficulties in detecting weak signatures of transiting small planets against a roaring background. These complications include instrument noise and systematic effects as well as the intrinsic stellar variability of the subjects under scrutiny. In this paper we review recent developments in the Kepler transit search pipeline improving both the yield and reliability of detected transit signatures.

Many of the phenomena in light curves that represent noise can also trigger transit detection algorithms. The Kepler Mission has expended great effort in suppressing false positives from its planetary candidate catalogs. Over 18,000 transit-like signatures can be identified for a search across 4 years of data. Most of these signatures are artifacts, not planets. Vetting all such signatures historically takes several months' effort by many individuals. We describe the application of machine learning approaches for the automated vetting and production of planet candidate catalogs. These algorithms can improve the efficiency of the human vetting effort as well as quantifying the likelihood that each candidate is truly a planet. This information is crucial for obtaining valid planet occurrence rates. Machine learning approaches may prove to be critical to the success of future missions such as TESS and PLATO.

We propose a low-contrast pre-coronagraph that can provide additional dark-hole contrast to a main coronagraph.

The Kepler Mission, combined with ground based radial velocity (RV) follow-up, has revolutionized the observational constraints on sub-Neptune-size planet compositions. Kepler's unprecedentedly large and homogeneous samples of planets with both mass and radius constraints open the possibility of statistical studies of the underlying planet composition distribution. This presentation describes the application of hierarchical Bayesian models to constrain the underlying planet composition distribution from a sample of noisy mass-radius measurements. This approach represents a promising avenue toward a quantitative measurement of the amount of physical scatter in small planet compositions, the identification of planet sub-populations that may be tied to distinct formation pathways, and empirical constraints on the dominant compositional trends in the planet sample. Both the transit and radial velocity techniques are subject to selection effects, and approaches to mitigate the resulting biases will be addressed. In addition to distilling composition-distribution insights from the current sample of Kepler planets with RV masses, this framework may be used to optimize the target selection for future transiting planet RV follow-up surveys.

Conversely to the transit photometry and radial velocity methods, the astrometric discovery of exoplanets is still limited by the sensitivity of available instruments. Ground-based surveys are now sensitive to giant planets in orbit around nearby low-mass stars and brown dwarfs. In 2014, ESA's Gaia mission began its survey, which is expected to discover thousands of giant exoplanets by detecting the astrometric orbital motions of the host stars.

Three versions of significance measures or False Alarm Probabilities (FAPs) for periodogram peaks are presented and compared for sinusoidal and box-like signals, with specific application on large-scale surveys in mind.

We present the first statistical analysis of the exoplanet frequency using planets found by a microlensing survey rather than follow-up observations. We present an analysis of 2007-2012 MOA (Microlensing Observations in Astrophysics) survey data to derive the planet frequency as a function of the planet/star mass ratio, q and separation, s, relative to the Einstein radius. Our sample includes 1472 microlensing events, including 22 planetary events and 1 ambiguous event with possible planetary and stellar binary solutions. The detection efficiency is calculated for each event and we employ a Bayesian analysis to deal with the ambiguous event. A broken power law model is used to fit the mass ratio function and we find a break and likely peak at q ~1.0−4.