PyStan provides a Python interface to Stan, a package for Bayesian inference using the No-U-Turn sampler, a variant of Hamiltonian Monte Carlo.

For more information on Stan and its modeling language, see the Stan User's Guide and Reference Manual at http://mc-stan.org/.

• HTML documentation: https://pystan.readthedocs.org
• Issue tracker: https://github.com/stan-dev/pystan/issues
• Source code repository: https://github.com/stan-dev/pystan
• Stan: http://mc-stan.org/
• Stan User's Guide and Reference Manual (pdf) available at http://mc-stan.org

• ArviZ: Exploratory analysis of Bayesian models with Python by @arviz-devs
• Prophet: Timeseries forecasting by @facebook
• Scikit-learn integration: pystan-sklearn by @rgerkin.
• BAMBI: BAyesian Model-Building Interface by @bambinos
• Jupyter tool: StanMagic by @Arvinds-ds
• Jupyter tool: JupyterStan by @janfreyberg

• PyMC: http://pymc-devs.github.io/pymc/

The development of PyStan3 with updated API can be found under stan-dev/pystan-next

Detailed installation instructions can be found in the doc/installation_beginner.md file.

Detailed installation instructions for Windows can be found in docs under PyStan on Windows

NumPy and Cython (version 0.22 or greater) are required. matplotlib is optional. ArviZ is recommended for visualization and analysis.

PyStan and the required packages may be installed from the Python Package Index using pip.

pip install pystan

Alternatively, if Cython (version 0.22 or greater) and NumPy are already available, PyStan may be installed from source with the following commands

git clone --recursive https://github.com/stan-dev/pystan.git
cd pystan
python setup.py install

To install latest development version user can also use pip

pip install git+https://github.com/stan-dev/pystan

If you encounter an ImportError after compiling from source, try changing out of the source directory before attempting import pystan. On Linux and OS X cd /tmp will work.

import pystan
import numpy as np
import matplotlib.pyplot as plt

schools_code = """
data {
    int<lower=0> J; // number of schools
    real y[J]; // estimated treatment effects
    real<lower=0> sigma[J]; // s.e. of effect estimates
}
parameters {
    real mu;
    real<lower=0> tau;
    real eta[J];
}
transformed parameters {
    real theta[J];
    for (j in 1:J)
        theta[j] = mu + tau * eta[j];
}
model {
    eta ~ normal(0, 1);
    y ~ normal(theta, sigma);
}
"""

schools_dat = {'J': 8,
               'y': [28,  8, -3,  7, -1,  1, 18, 12],
               'sigma': [15, 10, 16, 11,  9, 11, 10, 18]}

sm = pystan.StanModel(model_code=schools_code)
fit = sm.sampling(data=schools_dat, iter=1000, chains=4)

print(fit)

eta = fit.extract(permuted=True)['eta']
np.mean(eta, axis=0)

# if matplotlib is installed (optional, not required), a visual summary and
# traceplot are available
fit.plot()
plt.show()

# updated traceplot can be plotted with
import arviz as az
az.plot_trace(fit)