Bootstrapped Percentile Bootstrap

Common errors

• ModuleNotFoundError: No module named 'bootstrapped'

  cause The 'bootstrapped' package is not installed in the current Python environment.
  fix pip install bootstrapped

• TypeError: 'numpy.float64' object cannot be interpreted as an integer

  cause This error or similar type-related issues can occur due to strict type checking or implicit conversion changes in newer Python or NumPy versions, conflicting with the old code base of 'bootstrapped'.
  fix This often indicates a fundamental incompatibility with modern Python environments. It's recommended to try to replicate the original environment (e.g., Python 3.6, NumPy 1.12.x) or, preferably, migrate to `scipy.stats.bootstrap` or another actively maintained library.

• AttributeError: module 'bootstrapped' has no attribute 'bootstrap'

  cause The `bootstrap` function is located within the `bootstrapped.bootstrap` submodule, not directly in the top-level `bootstrapped` package.
  fix Use `import bootstrapped.bootstrap as bs` and then call `bs.bootstrap(...)`.

Warnings

• breaking This library has not been updated since September 2017. It is highly likely to have compatibility issues with newer versions of Python (e.g., Python 3.8+), NumPy, or SciPy, and may contain unpatched security vulnerabilities.
  Fix: It is strongly recommended to migrate to actively maintained modern statistical libraries such as `scipy.stats.bootstrap` (available since SciPy 1.6.0) or `arch` for bootstrap methods, which offer similar or enhanced functionality with active development and broader compatibility.
• gotcha The `bootstrapped` package's `num_iterations` parameter refers to the number of bootstrap resamples, not actual iterations of an algorithm. A higher number typically yields more stable results for confidence intervals.
  Fix: Ensure `num_iterations` is set to a sufficiently large number (e.g., 1000 or 10000) for reliable confidence intervals, understanding it controls the number of resamples rather than convergence of an iterative process.
• gotcha The library is maintained under `facebookincubator`, which implies it was an experimental project and not necessarily a fully supported product. This further emphasizes its abandoned status and lack of official support.
  Fix: Users should be aware that this project is not actively maintained or supported by Meta (Facebook) and should not be relied upon for production systems or critical analyses without significant independent validation.

Install

• pip install bootstrapped Install stable version

Imports

• bootstrap
  wrong:

  import bootstrapped as bs; bs.bootstrap(...)

  correct:

  import bootstrapped.bootstrap as bs

  `bootstrap` is a submodule, not directly available under the top-level package.
• stats_functions
  wrong:

  from bootstrapped import stats_functions

  correct:

  import bootstrapped.stats_functions as sf

  Common practice is to import as `sf` for convenience, but direct import also works.

Quickstart

This quickstart demonstrates how to generate sample data and then compute a 95% confidence interval for the mean using the `bootstrap` function from `bootstrapped.bootstrap` and the `mean` statistic function from `bootstrapped.stats_functions`.

import numpy as np
import bootstrapped.bootstrap as bs
import bootstrapped.stats_functions as sf

# Generate some data
data = np.random.normal(loc=100, scale=10, size=100)

# Calculate 95% confidence interval for the mean
lower_bound, upper_bound = bs.bootstrap(data, stat_func=sf.mean, alpha=0.05, num_iterations=1000)

print(f"Mean 95% CI: [{lower_bound:.2f}, {upper_bound:.2f}]")