Introduction

Stile is the Systematics Tests In LEnsing package. It’s designed to run systematics tests on lensing data; it’s especially designed for tests that you run on the data itself, such as PSF- galaxy shape correlation functions, as opposed to tests you run against an external data set, although you can do that too. The outputs of Stile are the systematic test results themselves; these can be compared against requirements that are set based on considerations and calculations that are external to Stile.

Installation

Stile is a pure python package, no compilation needed. You can install it using:

>>> python setup.py install

To run Stile, you must have:

• Python 2.7
• NumPy

We also recommend:

• TreeCorr, Mike Jarvis’s 2-point correlation function code. All of our correlation function tests involve calls to this package.
• PyFITS/Astropy to handle FITS tables and images. Stile can run on ASCII tables, but is much slower.
• matplotlib to generate plots.

While these are not requirements to import the Stile package and use its functionality, they are requirements for certain tests. None of the correlation function tests will run without TreeCorr, for example, and none of the whisker plots or scatter plots will run without matplotlib.

More dependencies may be added in the future.

Usage overview

The tests we’ve coded up live in the sys_tests module and we call them SysTests (for systematics tests, to distinguish from other kinds of tests). They’re all objects that you call to run a test. All have a plotting function as well (which doesn’t do anything if there’s nothing to plot), and possibly other helper functions. We’ve structured the code so that all the tests have the same kind of call signature: a dataset or datasets in a specific format described below, then any kwargs that control specific operation or plotting. The tests are defined in rough categories such as correlation functions, whisker plots, etc, and you can create either a flexible generic one or a use-specific version that requires fewer inputs on each run. For example, you could say:

>>> sys_test = stile.CorrelationFunctionSysTest('galaxy shear')

and then that object would only perform galaxy shear-shear correlations. The advantage is that the call signature is simple, so you don’t have to remember very much:

>>> corr_func = sys_test(galaxy_data)

Alternately, you could say:

>>> sys_test = stile.CorrelationFunctionSysTest()

and then you have an object you can call to make lots of different correlations. For example,

>>> corr_func = sys_test('gg', galaxy_data)

returns a shear-shear autocorrelation function of galaxies, or

>>> corr_func = sys_test('ng', lens_data, galaxy_data)

returns tangential shear around the data in lens_data. This more flexible version may be more useful for exploratory data studies, while the more specific version is best suited for automatic processing of data. (We have some tools in the works to aid automatic processing, but they are not yet complete.)

The data needs to be in a specific form, which is detailed further in the Data structure documentation.

Stile also contains some code to do simple binning of your data. This is described more in the Binning methods introduction. We also have a number of helper functions in File input and output, Stile utilities, and TreeCorr utilities, all described elsewhere in the documentation.