Kepler Astrophysical False Positive Probabilities Table Description
DR25 Table
This table contains the results of running an automated astrophysical false-positive-probability (FPP) calculating procedure on all transit signals identified in the DR25 KOIs Catalog. This procedure was first documented in Morton et al. (2012) and has evolved since then, with the most current description being in Morton et al. (2016). The astrophysical false positive scenarios considered by these calculations are blended and unblended eclipsing binary stars, where the eclipsing binary may be physically associated with the primary target or may be a chance-alignment with the primary target. This work does not consider a planet orbiting an unresolved companion to be a false positive scenario.
The code that implements this procedure is the Python module vespa, which is freely available online at https://github.com/timothydmorton/vespa. For completeness, we note that the version of vespa used to produce these results does is git commit 5157f60 rather than any officially tagged release of the code.)
A key part of these calculations involves estimating the posterior probability distributions of the physical parameters of the transit host stars, which are modeled as single, binary, and triple systems and constrained by both broad-band photometry and spectroscopy (when available). The vespa code uses the isochrones module (git commit 97061ad) to accomplish this, and the summary results of the single-star fits are presented in this table alongside the FPP results. Missing values are a result of failure modes as discussed in Section 4.7 ("Failure Modes") of Morton et al. (2016).
For this release, it's important to note that models have not been constrained by asteroseismic star properties due to an issue parsing the provenance flags. There are 233 KOIs associated with 184 unique host stars with an asteroseismic provenance (as indicated by entries beginning with the string "AST" in the column named logg_prov in the DR25 star properties catalog). Some of these host stars are giants misclassified as main sequence stars by vespa.
These FPP calculations represent an additional level of analysis that might help determine which KOIs are more likely to be planets and which are more likely to be astrophysical false positives. The results of this analysis have not been incorporated into any of the other Kepler data products. Consequently, they have not influenced the robovetter dispositions provided in the DR25 KOI table (Thompson et al. 2017, in preparation). These results help address the overall reliability of the catalog by quantifying the residual astrophysical false positive rate passing through the robovetter. It is possible that one or more key input assumptions of the FPP calculation are violated (e.g., the "exclusion radius" within which a blended eclipsing binary false positive might reside is under- or over-estimated). Users should proceed with caution when utilizing these statistics for individual candidates.
As discussed in Section 4.2 of Morton et al. (2016), FPP values are strictly reliable only for KOIs that have already strongly passed the other Kepler vetting tests and those that are not indicated as having poor fits to all the proposed hypotheses. FPPs should be reliable for KOIs passing the following criteria:
- KOI is dispositioned as CANDIDATE or CONFIRMED at the NASA Exoplanet Archive.
- KOI has MES > 10 indicating that the signal is unlikely to be caused by systematic noise in the light curve, and
- The positional probability indicates > 0.99 probability of the transit being on the target star according to Bryson & Morton (2017); in addition, the positional probability score should be higher than 0.3 (indicating a reliable result).
The input properties used in these calculations were pulled from the NASA Exoplanet Archive in April 2017. Since that time, the properties of 77 KOIs (plain text file) have been updated. The results in the FPP table do not reflect those changes.
It is important to note that the false positive probability (fpp_prob) quantifies astrophysical sources of false positives only. It does not include the probability that a transit signature is due to an instrumental artifact (though fpp_score should help to flag potential issues of this nature).
DR24 Table
This table contains the results of running an automated transit false-positive-probability (FPP) calculating procedure on all KOIs. This procedure was first documented in Morton (2012) and has evolved slightly since then, with the most current description being in Montet et al. (2015). The code used to implement this procedure is the Python module vespa, which is freely available online at https://github.com/timothydmorton/vespa.
A key part of these calculations involves estimating the posterior probability distributions of the physical parameters of the transit host stars, which are modeled as single, binary, and triple systems and constrained by both broad-band photometry and spectroscopy (when available). The vespa code uses the isochrones module to accomplish this, and the summary results of the single-star fits are presented in this table alongside the FPP results.
These FPP calculations represent an additional level of analysis that might help determine which KOIs are planets and which are false positives. The results of this analysis have not yet been incorporated into any of the other Kepler data products. Consequently, they have not influenced the robovetter dispositions provided in the Q1-Q17 DR24 KOI table (Coughlin et al. 2015) or the autovetter dispositions provided in the Q1-Q17 DR24 TCE table (Catanzarite et al., 2015). Hence, these results could address the reliability of these catalogs by identifying residual false positives. Of course, it is also possible that one of the key assumptions of the FPP calculation is violated (e.g., the "exclusion radius" inside of which a blended eclipsing binary false positive might live could be underestimated) or the transit signature is an instrumental artifact that is not properly modeled by this analysis.
Last updated: 20 September 2017
