Marshall is a tool for detecting artefacts masquerading as transits of small, long period planets in Kepler data. It is described in detail in Mullally et al. (2015, submitted to ApJ)

This code is published primarily as a record of how the results in that paper were produced, and running this package depends on code not yet made available. However, the core of the algorithm is self contained, even if the details of how I ran it on Kepler data are incomplete.

The exact implementation used in the paper is stored in tags/q1q17dr24. The trunk version should be easier to read.

The core of the algorithm is stored in the Trifit class in [trifit.py]. This class should be easily adaptable to any transit timeseries. The only function that assumes the input data is Kepler data is Trifit.isNotEnoughPointsInTransit(), which assumes it knows the exposure time of the input data. The Trifit class takes as input

• x An array of times of input data points
• y An array of flux values of input data points
• s An array of 1 sigma uncertainties on the flux values
• t0 The time of the midpoint of the proposed transit
• depth_frac The measured depth of the proposed transit in fractional units, where a depth of 1e-6 means a depth of 1ppm
• duration_days Duration of proposed transit
• ingress_days Ingress time of proposed transit

The method getTransitFavourability() returns a floating point number. Returns values > +10 indicate that the proposed event is more likely an artefact than a transit.

Considerable
[fitPipeline.py] and [marshall.py] are specific to Kepler data, and the particular problems I want to solve. They depend on a lot of functionality included in modules from a not-yet-available package, and will not work on your machine