boots 0.1.0

A tiny statistical bootstraping library.

Boots - A Tiny Bootstrapping Library

This is a tiny library for doing bootstrap sampling and estimating. It pulls together various tricks to make the process as fast and painless as possible. The tricks included are:

• Parallel execution with joblib
• The Bayesian bootstrap with two-level sampling.
• The Vose method for fast weighted sampling with replacement

Install

pip install boots

Example

import numpy as np

x = np.random.pareto(2, 100)

samples = bootstrap(
    data=x,
    statistic=np.median,
    n_iterations=1000,
    seed=1234,
    n_jobs=-1
)

# bayesian two-level w/ 4 parallel jobs
samples = bootstrap(
    data=x,
    statistic=np.median, 
    n_iterations=1000, 
    seed=1234, 
    n_jobs=4, 
    bayesian=True
)

# do something with it
import pandas as pd
posterior = pd.Series(samples)
posterior.describe(percentiles=[0.025, 0.5, 0.975])