Installing statsmodels

The easiest way to install statsmodels is to install it as part of the Anaconda distribution, a cross-platform distribution for data analysis and scientific computing. This is the recommended installation method for most users.

Instructions for installing from PyPI, source or a development version are also provided.

Python Support

statsmodels supports Python 3.8, 3.9, and 3.10.

Anaconda

statsmodels is available through conda provided by Anaconda. The latest release can be installed using:

conda install -c conda-forge statsmodels

PyPI (pip)

To obtain the latest released version of statsmodels using pip:

python -m pip install statsmodels

Follow this link to our PyPI page to directly download wheels or source.

For Windows users, unofficial recent binaries (wheels) are occasionally available here.

Obtaining the Source

We do not release very often but the main branch of our source code is usually fine for everyday use. You can get the latest source from our github repository. Or if you have git installed:

git clone git://github.com/statsmodels/statsmodels.git

If you want to keep up to date with the source on github just periodically do:

git pull

in the statsmodels directory.

Installation from Source

You will need a C compiler installed to build statsmodels. If you are building from the github source and not a source release, then you will also need Cython. You can follow the instructions below to get a C compiler setup for Windows.

If your system is already set up with pip, a compiler, and git, you can try:

python -m pip install git+https://github.com/statsmodels/statsmodels

If you do not have git installed or want to do the installation more manually, you can also type:

python -m pip install .

statsmodels can also be installed in develop mode which installs statsmodels into the current python environment in-place. The advantage of this is that edited modules will immediately be re-interpreted when the python interpreter restarts without having to re-install statsmodels.

python -m pip install -e .

It is usually recommended to use the --no-build-isolation to speed up the build process.

Compilers

Linux

If you are using Linux, we assume that you are savvy enough to install gcc on your own. More than likely, it is already installed.

Windows

It is strongly recommended to use 64-bit Python if possible.

Getting the right compiler is especially confusing for Windows users. Over time, Python has been built using a variety of different Windows C compilers. This guide should help clarify which version of Python uses which compiler by default.

macOS ^^^

Installing statsmodels on macOS requires installing gcc which provides a suitable C compiler. We recommend installing Xcode and the Command Line Tools, which can be done through the following command:

xcode-select --install

Dependencies

The current minimum dependencies are:

Cython is required to build from a git checkout but not to run or install from PyPI:

  • Cython >= 3.0.10 is required to build the code from github but not from a source distribution.

Given the long release cycle, statsmodels follows a loose time-based policy for dependencies: minimal dependencies are lagged about one and a half to two years. Our next planned update of minimum versions is expected in the first half of 2020.

Optional Dependencies

  • cvxopt is required for regularized fitting of some models.

  • Matplotlib >= 3 is needed for plotting functions and running many of the examples.

  • If installed, X-12-ARIMA or X-13ARIMA-SEATS can be used for time-series analysis.

  • pytest is required to run the test suite.

  • IPython >= 6.0 is required to build the docs locally or to use the notebooks.

  • joblib >= 1.0can be used to accelerate distributed estimation for certain models.

  • jupyter is needed to run the notebooks.

The optional dependencies can be installed along with statsmodels by modifying the installation command:

python -m pip install statsmodels[extras]

where <extras> is a comma-separated list of extras to install (build, develop, docs).


Last update: Oct 29, 2024