phdu 2.2b2

Automatically store/load data in a tidy, efficient way. Includes functions for data visualization and analysis.

PhD-utils

For people that have to compute and store a large variety of data and/or perform statistical inference.

Keep your files tidy!

Don't spend time creating directories, deciding filenames, saving, loading, etc. Decorators savefig & savedata will do it for you with optimal compression. More info at the tidypath repository.

Estimate confidence intervals

The module phdu.resample allows calls to the resample R package.

• Provides CI and permutation tests.
• CIs can account narrowness bias, skewness and other errors in CI estimation, as indicated in the article
• Alternatively, use phdu.stats.bootstrap for numba-accelerated computation (does not call resample).

Bootstrap-based power analysis.

Calculate the power for accepting H0 and estimate the needed sample size. Function power_analysis in phdu.stats.bootstrap follows Efron-Tshibirani: An introduction to the bootstrap, p. 381-384.

Numba-accelerated permutation tests

Module phdu.stats.tests.permutation.

• Permutation tests for any statistic.
• Includes paired and block cases.

Demo

Please check the example notebook.

Documentation

Github pages

Install

• For the R compatible installation first install R:

  conda install -c conda-forge r r-essentials r-base

• Install with dependencies:

  pip install phdu[dependencies]

  Where dependencies can be base (recommended), all, r (needed for resample to work), statsmodels, matplotlib or plotly.