SALib Version 1.1.0

New Features

• Refactored Method of Morris so the Ruano et al. local approach is default

Bug Fixes

• Inputs to morris.analyze are provided as floats
• Removed calls to standard random library as inconsistent between Python 2 & 3
• First row in Sobol sequences should be zero, not empty

Documentation

• Added a Code of Conduct
• Added DAETools, BCMD and others to citations - thanks for using SALib!
• Removed misleading keyword arguments in docs and readme examples
• Updated documentation for Method of Morris following refactor
• Improved existing documentation where lacking e.g. for fractional factorial
  method

Development Features

• Implemented automatic deployment to PyPi
• Fixed a bug preventing automatic deployment to PyPi upon tagging a branch
• Removed postgres from travis config

Stable version with fixes to documentation and testing suggested by reviewers.

Also some other small bugfixes.

Release of our stable version of SALIB to coincide with an submission to JOSS:

• Added a paper for submission to the Journal of Open-source Software
• Updated back-end for documentation on read-the-docs
• Updated the back-end for version introspection using PyScaffold, rather than versioneer
• Updated the Travis-CI scripts
• Moved the tests out of the SALib package and migrated to using pytest

This release adds improved sampling for the Morris method (thanks to @JoerivanEngelen ) and group sampling/analysis for the Sobol method (thanks to @calvinwhealton ).

@calvinwhealton has also added non-uniform distributions to the Sobol sampling. This will be a baseline for adding this to the other methods in the future.

Also includes several minor bug fixes.

• @dhadka has kindly contributed a wealth of documentation to the project, including doc strings in every module
• We no longer test for numpy <1.8.0 and matplotlib < 1.4.3, and these requirements are implemented in a new setup script.

New option to run analyze.sobol function in parallel using multiprocessing

This release does not contain any new functionality, but SALib now is citable using a Digital Object Identifier (DOI), which can be found in the readme.

Some minor updates are included:

• morris: sigma has been removed from the grouped-morris results and plots, replaced by mu_star_conf - a bootstrapped confidence interval. Mu_star_conf is not equivalent to sigma when used in the non-grouped method of morris, but its all we have when using groups.
• some minor updates to the tests in the plotting module

• Set up to include and test plotting functions
• Specific plotting functions for Morris
• Fractional Factorial SA from Saltelli et al.
• Repo transferred to SALib organization, update setup and URLs
• Small bugfixes

• Vectorized bootstrap calculations for Morris and Sobol
• Optional trajectory optimization with Gurobi, and tests for it
• Several minor bugfixes

Starting with v0.5, SALib is released under the MIT license.

Better Python API without requiring file read/write to the OS. Consistent functional API to sampling methods so that they return numpy matrices. Analysis methods now accept numpy matrices instead of data file names. This does not change the CLI at all, but makes it much easier to use from native Python.

Also expanded tests for regression and the Sobol method.

Improvements to Morris sampling and analysis methods, some bugfixes to make consistent with previous versions of the methods.

Improvements to Morris sampling methods (support for group sampling, and optimized trajectories). Much better test coverage, and fixed Python 3 compatibility.

First numbered release. Contains reasonably well-tested versions of the Sobol, Morris, and FAST methods. Also contains newer additions of DGSM and delta methods which are not as well-tested yet. Contains setup.py for installation.