Metadata-Version: 2.0
Name: selectinf
Version: 0.1.3
Summary: Testing a fixed value of lambda
Home-page: http://github.org/jonathan.taylor/selective-inference
Author: fixed_lambda developers
Author-email: 
Maintainer: Jonathan Taylor
Maintainer-email: 
License: BSD license
Platform: OS Independent
Classifier: Development Status :: 3 - Alpha
Classifier: Environment :: Console
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: BSD License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python
Classifier: Topic :: Scientific/Engineering
Requires: numpy (>=1.7.1)
Requires: scipy (>=0.9)
Requires: mpmath (>=0.18)
Requires: pyinter
Provides: fixed_lambda
Provides-Extra: test
Provides-Extra: doc
Requires-Dist: Cython (>=0.21)
Provides-Extra: doc
Requires-Dist: Sphinx (>=1.0); extra == 'doc'
Provides-Extra: test
Requires-Dist: nose (>=0.10.1); extra == 'test'


The selection project
=====================

This project contains software for selective inference, with emphasis on
selective inference in regression.

Some key references
-------------------

-  ``A significance test for the lasso``: http://arxiv.org/abs/1301.7161
-  ``Tests in adaptive regression via the Kac-Rice formula``:
   http://arxiv.org/abs/1308.3020
-  ``Post-selection adaptive inference for Least Angle Regression and the Lasso``:
   http://arxiv.org/abs/1401.3889
-  ``Exact post-selection inference with the lasso``:
   http://arxiv.org/abs/1311.6238
-  ``Exact Post Model Selection Inference for Marginal Screening``:
   http://arxiv.org/abs/1402.5596

Install
-------

.. code:: python

   git submodule init # travis_tools and C-software
   git submodule update
   pip install -r requirements.txt
   python setup.py install

Potential speedups
------------------

-  We can condition on “parts” of each draw of the sampler, in
   particular if we condition on the projection of the rejection
   ``sample - center`` onto direction then resampling on the ray can be
   sped up for some things like LASSO. Could be some cost in power.

-  Learning a higher dimensional function can perhaps save some time –
   proper conditioning has to be checked.



