The code used for generating the animations and data sets in the paper. The main purpose of this code is to provide a reference for performing these simulations, and ultimately has the goal of being a usable Python package for generating test data sets.
Currently this project is not available in any package repository like PyPI or Anaconda, so you will have to install from source:
$ git clone https://github.com/jmatejka/same-stats-different-graphs
$ cd same-stats-different-graphs
$ python setup.py install
This should download all dependencies, but just in case it's difficult to get the dependencies installed from PyPI, the packages needed for installation are:
setuptoolspandasseabornmatplotlibnumpyscipypytweeningtqdmdocopt
No specific tests have been performed yet, but this code should work fine with both Python 2 and Python 3.
Currently the easiest way to use this package is from the command line tool set
up by running python setup.py install, or by running the module directly with
python -m samestats:
$ python -m samestats run -h
Usage:
samestats run <shape_start> <shape_end> [<iters>][<decimals>][<frames>]
Right now we are working on getting the code cleaned up, documented, some features added, and tests written. Just the basics for getting a package ready for prime time. Once the code is in a pretty good state, we want to get it uploaded to PyPI and Anaconda, if possible, and the documentation uploaded to readthedocs.io.
We may also look into providing an easy-to-use API for generating these types of data sets from your own data, with the goal of making it easy to generate good test data for use in unit tests.
Another goal is to provide an API for "anonymizing" data sets. If the desired statistical properties can be preserved while adding noise to the underlying values, this could be a useful technique for protecting the privacy of study participants or users when publishing a dataset online.