PaperlayerView exact revision on arXiv

Source-linked AI summary

Kepler Data Validation I -- Architecture, Diagnostic Tests, and Data Products for Vetting Transiting Planet Candidates

Joseph D. Twicken, Joseph H. Catanzarite, Bruce D. Clarke, Forrest Girouard, Jon M. Jenkins, Todd C. Klaus, Jie Li, Sean D. McCauliff, Shawn E. Seader, Peter Tenenbaum, Bill Wohler, Stephen T. Bryson, Christopher J. Burke, Douglas A. Caldwell, Michael R. Haas, Christopher E. Henze, Dwight T. Sanderfer

arXiv:1803.04526v2astro-ph.EPastro-ph.IM

TL;DR

Kepler Data Validation addressed the need to characterize transit-search detections and distinguish genuine candidates from instrumental or astrophysical false positives. The paper describes DV’s architecture, model-fitting and residual searches, diagnostic tests, and archive products, with the final catalog workflow supporting uniform human and automated vetting. Its key contribution is a community-accessible set of per-target and per-candidate products, while difference-image diagnostics remain constrained for saturated, very low-S/N, or rapidly variable sources.

  • Problem

    Kepler needed to vet highly sensitive transit-search detections against instrumental, astrophysical, and background-source false positives.

  • Method

    DV characterizes TCEs, searches residual light curves for additional planets, applies diagnostic tests, and exports reports and tabulated results.

  • Results

    DV produced uniform diagnostic metrics and archive products that enabled consumers to assess Pipeline candidates without DV itself ranking their likelihood of being genuine planets.

  • Takeaways & Limitations

    The products supported community vetting of Kepler Pipeline candidates and subsequent scientific use of vetted catalogs.

  • Takeaways & Limitations

    Difference images may be unreliable for saturated sources, very low-S/N cases, or stellar variability on timescales comparable to transit durations.

Abstract

from arXiv · show

The Kepler Mission was designed to identify and characterize transiting planets in the Kepler Field of View and to determine their occurrence rates. Emphasis was placed on identification of Earth-size planets orbiting in the Habitable Zone of their host stars. Science data were acquired for a period of four years. Long-cadence data with 29.4 min sampling were obtained for ~200,000 individual stellar targets in at least one observing quarter in the primary Kepler Mission. Light curves for target stars are extracted in the Kepler Science Data Processing Pipeline, and are searched for transiting planet signatures. A Threshold Crossing Event is generated in the transit search for targets where the transit detection threshold is exceeded and transit consistency checks are satisfied. These targets are subjected to further scrutiny in the Data Validation (DV) component of the Pipeline. Transiting planet candidates are characterized in DV, and light curves are searched for additional planets after transit signatures are modeled and removed. A suite of diagnostic tests is performed on all candidates to aid in discrimination between genuine transiting planets and instrumental or astrophysical false positives. Data products are generated per target and planet candidate to document and display transiting planet model fit and diagnostic test results. These products are exported to the Exoplanet Archive at the NASA Exoplanet Science Institute, and are available to the community. We describe the DV architecture and diagnostic tests, and provide a brief overview of the data products. Transiting planet modeling and the search for multiple planets on individual targets are described in a companion paper. The final revision of the Kepler Pipeline code base is available to the general public through GitHub. The Kepler Pipeline has also been modified to support the TESS Mission which will commence in 2018.

1. Introduction

The Kepler Pipeline processed long-cadence observations through calibrated photometry, transit searches, and Data Validation to characterize and vet transiting planet candidates. Its design supported occurrence-rate studies while leaving candidate classification to human experts.

  • Kepler Mission: Kepler collected 29.4-minute long-cadence data for approximately 200,000 targets during the primary mission.Science operations emphasized detecting and characterizing Earth-size planets in the habitable zones of Sun-like stars.
  • Catalog Context: The final Q1–Q17 catalog contained 4034 vetted transiting planet candidates, including 219 absent from an earlier Kepler catalog.Vetted candidates were KOIs classified as Planet Candidates rather than False Positives.
  • Pipeline Architecture: The Pipeline calibrated pixels, extracted photometry and centroids, corrected systematic errors, searched for transits, and generated Threshold Crossing Events.These stages were implemented through CAL, PA, PDC, and TPS before Data Validation.
  • Data Validation: Data Validation characterized TCEs, searched for additional planets after transit removal, and applied uniform diagnostic tests to aid candidate assessment.DV was not tasked with rating candidates as bona fide planets; those decisions were left to human experts.

2. Pipeline Data Validation

Data Validation independently processes TCE-producing targets by fitting transit models, searching residual light curves for additional candidates, running diagnostics, and exporting structured vetting products. The DR25 implementation operated at substantial computational scale under per-target resource limits.

  • Front-End Processing: DV independently processed each target that produced a TCE, fitting transiting-planet models before searching residual light curves for additional candidates.The fit used systematic-error-corrected, quarter-stitched light curves, and DV also supported odd-even, trapezoidal, and reduced-parameter fits.
  • Diagnostic Processing: After modeling and multiple-planet searches, DV sequentially applied configurable diagnostic tests to discriminate genuine transiting planets from false positives.The tests could be enabled or disabled individually and differed in computational cost and expected return.
  • Archive Products: DV generated four product types: target-level reports, TCE-level summaries, target time-series FITS files, and run-level XML results.The Archive component generated the time-series and XML products after DV processing, and products were exported for community access.
  • DR25 Execution: 198,707 targets were searched in DR25, 17,230 produced TCEs, and the median DV runtime was 9.47 hours under a 45-hour timeout.The fitter and multiple-planet search received 36 hours of the allocated limit.

3. Diagnostic Tests

DV diagnostic tests produced metrics for distinguishing bona fide transiting planets from false positive detections. The test suite addressed instrumental, astrophysical, and background-source causes of transit-like signals.

  • Diagnostic Test Scope: DV applied a diagnostic suite to every Pipeline-identified planet candidate, including candidates found during DV’s multiple-planet searches.The tests were intended to support discrimination between bona fide planets and false positive detections.

3.1. Weak Secondary Test

The weak secondary test searches for secondary eclipse signatures by measuring phase-dependent Multiple Event Statistics after removing primary eclipses. Its examples show how secondary signals distinguish background eclipsing binaries from planetary occultations, while remaining non-exclusive evidence.

  • Test operation: The weak secondary test computes Multiple Event Statistics versus orbital phase after removing the primary transit signature.The search uses the TCE’s orbital period and trial pulse duration; the resulting phase vector identifies maximum and minimum secondary events.
  • Interpretive caveat: Matching-period TCEs or secondary eclipses do not by themselves establish an eclipsing binary because short-period planets can produce reflected-light or thermal occultations.DV models secondary events and compares inferred albedo and effective temperature with physically relevant expectations.
  • Diagnostic examples: 11.4σ secondary MES at phase 9.222 days exposed KOI 140 as a background eclipsing binary offset by ∼6 arcsec.Because the signal exceeded the 7.1σ threshold, the multiple-planet search generated a second TCE with MES = 11.9σ at the same 19.978-day period.
  • Diagnostic examples: 5.1σ secondary MES at phase 0.781 days remained below the 7.1σ threshold for KOI 2887, so no second TCE was produced.The normalized phase, 0.781/1.569 = 0.50, and pixel inspection nevertheless indicated a circular eclipsing binary in the background source.
  • Planetary versus binary interpretation: 60.8 ± 1.65 ppm secondary depth, albedo 0.167 ± 0.012, and temperature 2026 ± 31 K supported a planetary occultation for HAT-P-7b.The reported modeling was consistent with reflected light or thermal occultation from a giant planet.
  • Planetary versus binary interpretation: Albedo 7.82 ± 1.73 and temperature 2632±72 K, 16.9σ above the 1018±62 K equilibrium temperature, rejected a giant-planet occultation for KOI 6167.01.Its secondary eclipse depth was 3143 ± 61 ppm, and the object is identified as a 3.9-day eclipsing binary.

3.2. Rolling Band Contamination Diagnostic

The rolling band contamination diagnostic identifies transits coinciding with temperature-dependent image artifacts and summarizes their severity. Examples show that substantial coincidence can undermine candidate credibility, although some artifacts may be present without invalidating a planet detection.

  • Diagnostic construction: Severity levels range from 0 (low) to 4 (high), and the diagnostic reports both transit counts and fractions for each level.The resulting table is provided for every DV planet candidate.
  • Diagnostic construction: The diagnostic counts observed transits coincident with rolling band artifacts at each severity level, using the pulse duration closest to the transit duration.Severity levels are derived from target-specific time series generated from readout-channel and CCD-row artifact metrics.
  • Example: HAT-P-7b: For HAT-P-7b, 529 of 584 transits had severity level 0, while 46, five, one, and three transits had levels 1, 2, 3, and 4 respectively.The example uses the 3.0 hr artifact pulse duration, closest to the 4.04 hr modeled transit duration.
  • Scope boundary: Target-specific severity levels are undefined for missing or anomalous cadences and for quarters in which Dynablack was not run, including Q1 and Q17.These gaps constrain when the contamination diagnostic can be evaluated.
  • Interpretation: Rolling-band contamination warrants particular attention when a significant fraction of transits coincide with nonzero severity, especially for long-period TCEs with few observed transits.Some HAT-P-7b transits coincided with artifacts, but most were assigned severity level 0.
  • Example: long-period candidates: Three KIC 8373837 TCEs had rolling-band coincidence fractions of 3/4, 2/4, and 4/4, and none represented credible transiting planets.Their events occurred in quarters when the target was observed on a known artifact channel.

3.3. Eclipsing Binary Discrimination Tests

DV uses statistical comparisons of transit depths, epochs, and candidate periods to discriminate eclipsing binaries from planetary transit sources. The tests identify known false-positive patterns while retaining caveats for short-period planetary occultations.

  • Odd/even transit tests: DV fits odd and even transit sequences separately and tests whether their depths are statistically equal.Genuine planets are expected to have consistent odd and even transit depths, subject to spacecraft-roll effects.
  • Odd/even transit tests: Odd/even epoch comparisons test whether timing differs from one-half of the fitted orbital period, flagging possible slightly eccentric binaries.This diagnostic is computed even though the single-TCE eccentric-binary case almost never occurs.
  • Shorter/longer period tests: When multiple TCEs occur, DV compares each candidate’s period with the nearest shorter- and longer-period candidates using transit durations as period uncertainties.Statistical period equivalence commonly indicates separate TCEs from primary and secondary eclipses of a binary, though short-period planetary occultations can produce a similar signal.
  • Statistical formulation: The tests are implemented as χ2 hypothesis tests that support statistical comparison of independent measurements.The χ2 formulation uses a weighted mean and a χ2 distribution with N −1 degrees of freedom.
  • Examples: KOI 6996 shows a 7312 ± 35.5 ppm odd/even depth difference at 206σ, identifying a circular eclipsing binary detected at half its true period.For KOI 140, primary and secondary eclipses generated periods of 19.9782 and 19.9787 days, inconsistent with a planetary classification.

3.4. Difference Imaging and Centroid Offset Analysis

Difference imaging and centroid-offset analysis use spatial pixel information to locate transit sources and distinguish target-associated signals from background false positives. The diagnostics identify bona fide transits with insignificant offsets and expose background eclipsing binaries through large, significant displacements.

  • Difference Image Generation: Difference imaging compares mean in-transit and out-of-transit pixel fluxes to locate the source of a transit signature within or beyond the photometric aperture.Images are generated per target, candidate, and quarter when a planet-model or fallback trapezoidal fit is available and at least one clean transit is observed.
  • Centroid Offset Analysis: Centroid offsets compare difference-image and out-of-transit centroids, while KIC-referenced offsets provide an alternative subject to catalog errors and centroid bias.Quarterly offsets supply absolute and statistical measures of separation between the target and transit source.
  • Limitations: Difference images may be unreliable for saturated sources, very low S/N, short-timescale stellar variability, or overlapping transits retained to avoid losing a quarterly image.DV computes a quality metric, but Q1–Q17 DR25 could retain potentially misleading images when overlap would otherwise prevent construction.
  • Centroid Offset Analysis: Robustly averaged quarterly offsets improve localization accuracy and assess whether the transit source is statistically distinguishable from the target.Inverse-variance weighting emphasizes precise measurements, and robust averaging deemphasizes outliers.
  • Results: 0.1010±0.0819 arcsec (1.23σ) relative to the out-of-transit centroid and 0.1365 ± 0.0969 arcsec (1.41σ) relative to the KIC position supported Kepler-11e as a bona fide transiting planet.The target and difference-image centroid markers were closely spaced and statistically indistinguishable.
  • Results: 5.801 ± 0.073 arcsec (79.4σ) relative to the out-of-transit centroid identified KOI 140 as a background eclipsing-binary false positive.The difference-image centroid was nearly coincident with KIC 5130380, which is 2.5 magnitudes (10 times) fainter than the target.

3.5. Statistical Bootstrap

The statistical bootstrap estimates each TCE’s noise-only false-alarm probability from null single-event statistics and the observed multiple-event detection statistic. It yields extremely small probabilities for the example candidates, supporting their detection reliability within the stated statistical framework.

  • Purpose: The bootstrap estimates the probability that noise alone would produce a TCE with the same or larger multiple event detection statistic.This false-alarm probability is used to assess TCE reliability.
  • Bootstrap Inputs: Null SES time series are generated after removing identified transit events, and the series at the TCE’s trial pulse duration supplies the bootstrap input.Null statistics may be unavailable when the multiple-planet search halts after reaching its iteration limit.
  • Statistical Construction: The multiple-event statistic Z combines P single-event statistics, with C(p) and N(p) representing correlation and normalization statistics for the pth transit.The joint P-event distribution is formed from the single-event distribution and computed efficiently with Fourier transforms.
  • Statistical Construction: The joint distribution is collapsed into a one-dimensional Z histogram, whose upper-tail probabilities estimate false alarms above a given detection statistic.Typical histogram bins have width 0.1σ.
  • Results: 2.97x10−13 at MES = 7.7σ was estimated for Kepler-186f, while 6.27x10−17 at MES = 8.5σ was estimated for Kepler-62c.The Kepler-62c estimate used asymptotic extrapolation.
  • Results: 10−12 was the expected false-alarm probability for one statistical false positive under whitened Gaussian noise, motivating the 7.1σ Pipeline detection threshold.The estimate reflects approximately 10^12 statistical tests across the four-year search.

3.6. Centroid Motion Test

The centroid motion test assesses whether a candidate’s centroid shifts correlate with its transit signal and estimates the transit source’s offset from the target. It combines coordinate-specific detection statistics, significance testing, and source-location diagnostics, with illustrative results for a false positive and a confirmed planet.

  • Preprocessing: Flux-weighted centroids are converted from CCD coordinates to right ascension and declination, then cotrended to remove spacecraft and motion-related systematics.The centroid aperture includes the optimal photometric aperture plus one halo ring, and coordinate conversion uses inverted PA motion polynomials.
  • Test objectives: The test evaluates transit–centroid correlation and whether the inferred transit source is offset from the target, while allowing crowding or imperfect background removal to mimic correlated motion.Low correlation makes a background source unlikely; significant correlation can indicate a background source but can also arise from target-aperture effects.
  • Detection statistic: Transit-model light curves are jointly fitted in amplitude to the two centroid time series in an iteratively whitened domain, with other candidates’ transit signatures removed before each statistic is computed.Detection statistics are computed separately for right ascension and declination, then combined for each planet candidate.
  • Statistical significance: The total centroid-motion statistic sums the squared coordinate statistics, follows a χ2 distribution with two degrees of freedom, and is reported with a p-value when transit-model fitting and iterative convergence succeed.The p-value is the probability of obtaining a statistic at least as large from random centroid fluctuations without correlated transit-induced motion.
  • Examples: For KOI 140, the estimated source offset was 19.8 arcsec (146σ), overestimating the believed 5.8 arcsec background-eclipsing-binary offset because the source often fell outside the aperture.The aperture-related depth underestimation contributed to the discrepancy.
  • Examples: For Kepler-62f, the peak centroid shift was 0.294 ± 0.338 mas and the total statistic was 2.57 with p = 0.28, indicating no statistically significant centroid motion.The estimated source offset was 1.08 arcsec (1.50σ).

3.7. Optical Ghost Diagnostic Test

The optical ghost diagnostic compares transit-model correlations in core and halo apertures to distinguish target-centered transit signals from distributed contamination. In DR25, the expected ordering held for 98.1% of classified “golden” planet candidates, while representative ghost and background-eclipsing-binary cases showed the reverse ordering.

  • Test architecture: The test correlates whitened core and halo aperture flux time series with the candidate’s whitened transit model after removing other candidates’ transit signatures.Core and halo statistics are computed from calibrated, corrected, detrended, stitched, and gap-filled aperture time series.
  • Test architecture: The test runs only when sufficient Data Validation time remains for computation and subsequent report generation.It is computationally intensive and may be omitted when runtime is constrained.
  • Interpretation: A target-origin transit is expected to produce a larger core than halo correlation because targets are generally centered in their photometric apertures.The DR25 run produced this ordering for 3291 of 3348 classified “golden” KOIs, or 98.1%.
  • Interpretation: Broad optical ghosts are expected to produce a larger halo than core correlation because halo flux is subtracted from core flux before the core statistic is computed.The same ordering can also occur for background eclipsing binaries outside the target’s quarterly photometric apertures.
  • Representative results: KOI 3900.01 illustrates the diagnostic: its core time series was not highly correlated with the transit model, and the source was identified as an antipodal ghost.The ghost was attributed to reflection of a bright eclipsing binary from the Schmidt corrector plate.
  • Representative results: Representative results showed core-dominant correlations for confirmed-planet KOIs and halo-dominant correlations for optical-ghost and background-eclipsing-binary false positives.The examples include KOIs 157.03, 701.04, 571.05, 3900.01, 4718.01, and 140.01.

4. KOI Matching

KOI matching links known KOI ephemerides to Data Validation TCEs for archive users without using prior KOI knowledge to guide transit detection or validation. Matching uses correlations between rectangular transit time series, with ambiguity and non-linear-ephemeris cases limiting reported matches.

  • Purpose and inputs: KOI matching is performed after validation by comparing known KOI ephemerides with TCE ephemerides at target and planet levels.The results are reported in archive products for consumer use.
  • Matching method: Target-level matches compare integer KIC IDs, while planet-level matches correlate rectangular transit series generated from periods, first-transit epochs, and durations.The correlation compares each known KOI associated with a target against its TCEs.
  • Matching method: A KOI and TCE match when their ephemeris correlation is at least the configurable threshold, typically 0.75.The threshold was selected so reported matches are highly likely while allowing low-level ephemeris discrepancies.
  • Caveats: No match is reported when one KOI exceeds threshold for multiple TCEs or one TCE exceeds threshold for multiple KOIs.This ambiguity occurs for duplicate KOIs such as 1101.01/1101.02 and 2768.01/2768.03.
  • DR25 results: Among 3402 “golden” KOIs, 3354 received matches at threshold or better, and 92.0% of those matches had correlation coefficients above 0.9.Of the 48 remaining KOIs, 40 were recovered in the transit search without producing threshold-level ephemeris matches.
  • DR25 results: Matching failures arose from period-factor differences, incorrect or ambiguous periods, transit-timing variations, and eclipsing-binary or heartbeat-star ephemerides.Some systems lack a true linear ephemeris, limiting correlation-based matching.

5. Archive Products

Data Validation archive products summarize inputs, model fits, and diagnostic-test results for community vetting of transiting planet candidates. The final DR25 system generated four product types at target, candidate, and pipeline-run levels.

  • Purpose: Archive products document the information supplied to Data Validation and the results of transit model fits and diagnostic tests.They support community assessment rather than determining the likelihood that a TCE is a legitimate planet.
  • Product types: The four products are comprehensive target-level DV Reports, one-page TCE-level Report Summaries, target-level DV Time Series FITS files, and a pipeline-run DV XML file.The products summarize reports, candidate results, relevant time series, and tabulated results.

5.1. DV Report

The DV Report consolidates target, data, candidate, light-curve, and diagnostic information into accessible products for reviewing each target and its TCEs. Its visual summaries support rapid assessment while requiring caution about data-quality artifacts and red diagnostic flags.

  • Report structure: Each target with at least one TCE receives a comprehensive PDF DV Report, automatically generated and delivered to the NASA Exoplanet Archive.The report is organized into logical sections with tabs for locating target-specific results.
  • Report structure: The report summary tables document stellar parameters with uncertainties and provenance, quarter-specific data characteristics, and candidate properties including orbital and thermal quantities.Candidate entries also flag possible eclipsing binaries based on transit depth, in which case some transit model fits are omitted.
  • Flux time series: Quarter-stitched PDC light curves mark candidate transits and expose gaps, spacecraft events, noisy segments, and other conditions that can make detections suspect.These displays are particularly useful for candidates supported by relatively few transits.
  • Flux time series: Comparing quarterly PA/SAP light curves with quarter-stitched PDC curves can reveal post-PA processing problems, including degradation of short-period transit signatures.Transits visible in PA but absent from PDC are identified as a red flag.
  • Dashboards: Candidate dashboards summarize model-fit and diagnostic results using colors that distinguish nominal, borderline, questionable, and unavailable outcomes.A red region does not by itself invalidate a candidate because some genuine planets can trigger eclipsing-binary or centroid-motion diagnostics.
  • Diagnostics: Centroid cloud plots relate flux changes to right-ascension and declination centroid changes, making correlated in-transit motion directly visible while not excluding motion when correlation is absent.Rolling-band tables separately report transit coincidences with image artifacts by severity level.

5.1.7. Pixel Level Diagnostics

The report presents pixel-level, phase-folded, model-fit, and diagnostic-test products for each candidate, with supporting residual, odd/even, alert, and appendix information. Together these views expose fit quality and potential false-positive indicators without reducing interpretation to a single diagnostic.

  • Pixel-level diagnostics: Pixel-level diagnostics provide difference-image quality metrics, centroid offsets, and quarter-by-quarter image displays for each candidate.The products include both graphical and tabular representations, with quality thresholds and cadence information documented for each image.
  • Phase-folded views: Phase-folded light curves are shown in unwhitened and whitened domains, with transit markers supporting comparisons among candidates in multiple-TCE systems.Later releases also provide phase-folded views by quarter, observing season, and year after median detrending.
  • Candidate fit products: Each candidate section combines TCE parameters, transit-model fit results, quarterly light curves, phase-folded diagnostics, and reduced-parameter fit results.The whitened model is overlaid on phase-folded data, while markers distinguish emphasized from deemphasized or ignored points.
  • Diagnostic tests: Separate tables report weak-secondary, centroid-motion, eclipsing-binary, bootstrap, and optical-ghost diagnostic results for each candidate.These tests extend the model-fit displays with distinct indicators of possible nonplanetary explanations.
  • Supporting diagnostics: Appendices add robust-fit weights, residual histograms, and odd/even transit fits that support interpretation of fit irregularities and eclipsing-binary discrimination.The odd/even comparison uses the fitted depth difference relative to its uncertainty.
  • Alerts: Runtime alerts record timestamps, severity states, and messages identifying the target, candidate, and DV subcomponent associated with off-nominal conditions.DV alerts largely identify warning conditions rather than errors.

5.2. DV Report Summary

The one-page DV Report Summary condenses model-fit and selected diagnostic results for each TCE into a format used for rapid manual vetting. It combines visual transit views, tabulated parameters, diagnostic flags, stellar information, and processing metadata.

  • Summary purpose: A one-page PDF Report Summary is generated for each TCE and includes diagnostic figures plus tabulated model-fit and diagnostic-test results.It became a basis for TCERT triage and contributed to KOI promotion and classification activities.
  • Summary figures: The summary displays detrended quarter-stitched and phase-folded light curves, with additional focused views centered on the primary transit and strongest secondary eclipse.The phase-folded panel overlays the transit model on the full light curve.
  • Summary tables: A text box reports fitted and derived planet parameters, secondary-event parameters, uncertainties, and selected diagnostic results with significance values where applicable.Results statistically inconsistent with a planetary classification are highlighted in red.
  • Metadata and guidance: The summary also shows stellar parameters, KOI matching results, warnings for assumed Solar values, generation time, and the SOC code branch used by DV.A detailed guide for the DR25 version uses Kepler-186f as its case study.

5.3. DV Time Series

The DV Time Series file is a FITS data product for each long-cadence target with at least one TCE. It stores time-series information relevant to target processing and all associated TCEs.

  • File scope: The AR component generates one FITS DV Time Series file for each long-cadence target with at least one TCE.The file contains time-series data relevant to TPS/DV processing for the target and its associated TCEs.

6. Conclusion

The paper describes Data Validation as the Pipeline’s final component for characterizing candidates, finding additional planets, and supporting comprehensive vetting. It documents the DV architecture, diagnostic tests, and data products, while releasing the final code base for community use.

  • DV characterizes transiting planet candidates, searches for additional planets after removing modeled transit signatures, and performs diagnostic tests for candidate vetting.
  • The paper documents the DV architecture, diagnostic-test suite, and data products produced for Pipeline transiting planet candidates.
  • The final revision of the DV code base was released through GitHub for the benefit of the community.
Loading 1803.04526v2…