empiricaldist is a Python library that provides classes to represent empirical distributions -- that is, distributions based on data rather than mathematical functions. It includes four equivalent ways to represent a distribution: PMF (Probability Mass Function), CDF (Cumulative Distribution Function), Survival function and Hazard Function. It provides methods to convert from one representation to the others, and methods to perform a variety of operations.

This library is used extensively in Think Stats, Think Bayes, Elements of Data Science, and Think Complexity -- but it is intended to be a stand-alone library for general use, not just for my books.

Here's a quick example of how to use it:

from empiricaldist import Pmf

# Create a PMF object
pmf = Pmf.from_seq([1, 2, 2, 3, 5])

# Make the other representations
cdf = pmf.make_cdf()
surv = cdf.make_surv()
hazard = surv.make_hazard()

# Look up quantities
print(pmf(4))
print(cdf(4))
print(surv(4))
print(hazard(4))

# Cdf and Surv also provide inverse lookups
print(cdf.inverse(0.5))
print(surv.inverse(0.5))

For an introduction to the API, you can read this notebook or run it on Colab.

To install empiricaldist, use pip:

pip install empiricaldist

empiricaldist is available under the BSD 3-clause license. See the LICENSE file for more details.